Ch4 Numerical Approaches For The NVH Study of Electric and Hybrid Electric Vehicles

NVH Analysis Techniques for Design and
Optimization of Hybrid and Electric Vehicles

Chapter 4
Numerical Approaches for the NVH study of Electric and
Hybrid Electric vehicles
Paul Borza1 and Sofia Sanchez Mateo 2
1
Electronics and Computers Department, Transilvania University of Brasov,
Politehnicii Street 1, Brasov 500024, Romania, borzapn@unitbv.ro,
2
Mechanical Engineering and Energy Department, Miguel Hernandez University of
Elche, Av de la Universidad, S/N - 03202, Elche, Spain,
sofia.sanchez@graduado.umh.es,
Abstract. This chapter is dedicated to an overview of models and methodologies used for the
numerical analysis of noise, vibration, and harshness of the mechanical systems, with a focus to
EV’s. After a short general introduction, the annexed sections compile relevant studies
performed at TU1105 partner organisations with the aim to show some relevant and high-
potential recent developments in the field.
1 Introduction
Noise, Vibration & Harshness (NVH) describes the dynamic behaviour of vehicles in
the whole frequency range. It comprises the sensible low frequency vibrations of
structures, the sensible and audible range of perceived acoustical comfort (harshness),
as well as the audible high frequency range of noise limited by the perception ability
of the human ear (20 Hz – 20000 kHz).
Due to the complexity of the NVH attribute, typically automotive engineers rely on
the use of numerical models for NVH analysis and design. The most widely used
numerical methods are the Finite Element Method (FEM) and the Boundary Element
Method (BEM). Both of them can be used for structural and acoustic analysis but
typically, FEM is used for structural and bounded acoustic problems whereas BEM is
used for (unbounded) acoustical problems. They are mainly applicable for low
frequency analysis. For high-frequency analysis, other methodologies such as
Statistical Energy Analysis is often used. To analyse low-frequency mechanical
motion of vehicle systems (such as suspensions), other time-domain integration based
schemes such as Multi-Body analysis or 1D System Simulation can be deployed.
Fig. 1. Simulation methods and their application range Erreur ! Source du renvoi
introuvable.
2 Methodologies and numerical contributions to the study of

EVs
According to the objective of this chapter, several issues related to NVH numerical
methods have been studied and analysed in detail. If we look how these methods find
their application in the various challenges related to the NVH behaviour of (H)EV,
the Source-Transfer-Receiver approach, well known from experimental analysis,
modelling and refinement, can be adopted as well to identify and address the
numerical modelling challenges [2].
Fig. 2 shows the main topics of this chapter. Each input is presented as an individual
annex, according to the following structure:
4A. Vibro-acoustic metamaterial developments combining low-mass

structural solutions with high vibro-acoustic insulation performance.
4B. Development of equations of motion based on Lagrange
formulation, applied for the modal analysis of mechanical systems.
4C. Using of BEM simulation approaches for sound field prediction of
EVs and alerting systems.
4D. Equivalent material modelling of sandwich beam assemblies:
propagating and evanescent wave considerations.
4E. Modelling the NVH and thermal behaviour of electric motors using a
combination of electromagnetic FEM, 1D system simulation, structural FEM
and Acoustic BEM.
______________________________________________________________________________________
NVH Analysis Techniques for Design and Optimization of Hybrid and Electric Vehicles
214
Nuria Campillo-Davo and Ahmed Rassili, ISBN 978-3-8440-4356-3, Shaker Verlag Publications (2016)
4F. Using FEM and BEM methods to design an optimal warning sound
system to protect Vulnerable Road Users.
4G. Using FEM material modelling for the vibrational analysis of inertial
dampers of EVs, application to damage detection using experimental modal
identification
FEM modelling for

materials design and NVH
analysis-
Annexes 4A, 4D & 4G
Development Multyphysics
of numerical NVH
methods THE DEVELOPMENT OF SimulaTon of
applicable to NUMERICAL NVH Electric
NVH analysis - ANALYSIS TECHNIQUES Drivetrains -
Annex 4B APPLIED TO EVS Annex 4E
BEM approach to the

FEM/BEM approach predicTon of sound
to the design of EVs field produced by EVs
Warning Sound and alerTng systems -
systems - Annex 4F Annex 4C
Fig.2. Main aspects of the numerical NVH study of Electric and Hybrid Electric Vehicles
3 Conclusions
Some important achievements have been presented in this chapter, as well as

suggestions for future work. Concerning the contributions compiled and related to
experimental approaches for the NVH analysis of EVs and HEVs, the main
conclusions are:
- With the constant quest towards low mass vehicles, the need arises to develop and
model lightweight materials and structures with enhanced NVH behaviour. Annex
4A discusses the developments regarding acoustic resonant metamaterials at KU
______________________________________________________________________________________
215
Leuven [3, 4]. By exploiting inherent dynamic properties of integrated material
resonating elements, vibro-acoustic stopband behaviour can be created at low
frequencies. This allows to combine lightweight solutions with good NVH
properties and opens up possibilities to create dedicated solutions for a.o.
electrified vehicles to contain and shape noise and vibration patterns.
- Annex 4B shows a numerical formulation based on Lagrange equations, useful for
an optimized finite element analysis of vibration.
- Annex 4C addressed the potential of BEM for the prediction of the sound field
generated by EVs and HVs and integrated alerting systems (AVAS) in an urban
environment.
- Annex 4D studies the dynamic behaviour of a sandwich laminate structure. On the
basis of a practical example, the equivalent loss factor was defined and applied to
the equivalent material.
- The multiphysics simulation methodology presented in Annex 4E has been applied
to the specific problem of an SRM (Switched Reluctance Motor). More detailed
discussions can be found in [5 - 7]. Further refinement of the method, a better
integration with the EMC simulations and a validation of experimental test cases
is ongoing.
- The warning sound approach discussed in Annex 4F fits into the broader
discussion of protecting vulnerable road users against low-speed incidents by
electric Vehicles appearing unadvertedly. The FEM/BEM approach to validate the
loudspeaker configuration is discussed here. The actual definition and generation
of the sound and the pedestrian detection system are discussed elsewhere. [8 - 10].
- Annex 4G uses FEM for modelling the behaviour of composite materials that
inertial dampers are made of. Vibrational analysis of such models, combined with
experimental modal identification, is applied for damage detection.
4 References
[1] Bierman, J. W., NVH Calculation Technologies, ika, Aachen University, e-COST
TU 1105, Training School in Liege (2015)
[2] H. Van der Auweraer, K. Janssens, A Source-Transfer-Receiver Approach to
NVH Engineering of Hybrid/Electric Vehicles. SAE paper 2012-36-0646, Proc.
SAE NVH Brazil, Florianopolis, Brazil, 2012
[3] C. Claeys, B. Pluymers, P. Sas, W. Desmet. Design of a resonant metamaterial
based acoustic enclosure. ISMA2014 - USD2014. Leuven, Belgium, 15-17
September 2014 (art.nr. 561) (pp. 3351-3358).
[4] C. Claeys, E. Deckers, B. Pluymers, and W. Desmet, A lightweight vibro-acoustic
metamaterial demonstrator: Numerical and experimental investigation,
Mechanical Systems and Signal (2016), 70-71, 904-918.
[5] J. Anthonis, F. Santos, V. Croitorescu, H. Van der Auweraer, “Multiscale Thermal
and NVH Models for EV/HEV Integration of an SR-based Drivetrain”, Virtual
Powertrain Conference, Oct. 19-20, 2011, Pune, India
______________________________________________________________________________________
216
[6] F.L.M. Dos Santos, J. Anthonis, F. Naclerio, J.J.C. Gyselinck, H. Van der
Auweraer, and L.C.S. Goes, Multiphysics NVH Modeling: Simulation of a
Switched Reluctance Motor for an Electric Vehicle, IEEE Transactions on
Industrial Electronics, 61-1(2014), 469–76.
[7] C.T. Faria, F.L.M. Dos Santos, F. Chauvicourt, S. Orlando: Noise emissions on
switched reluctance motors: evaluation of different structural models, Procedings
of EVS28, Kintex, Korea, 2015
[8] J. J. García , J. M. Barrios , D. Quinn , G. Baudet ,T. Schmit , J. Chamard, “The
expected impact of eVADER technology on the safety of vulnerable road users”,
Proc. ISMA 2014, Noise and Vibration Conference (Leuven, Belgium, 15-17
September 2014
[9] H. Van der Auweraer, D. Sabbatini, E. Sana, K. Janssens, “Analysis of Electric
Vehicle Noise in View of VRU Safety”, in Proceedings ICSV18, Brazil, 10-14
July 2011
[10] Van Genechten B., Vansant K. & Berkhoff A.P., Simulation-based design of a
steerable acoustic warning device to increase (H)EV detectability while reducing
urban noise pollution In Proceedings of the TRA 2014, Paris, France, 14-17 April
2014
______________________________________________________________________________________
217
Annex 4A
Design and validation of a metamaterial acoustic
enclosure
Claus Claeys1, Elke Deckers, Lucas Van Belle, Alireza Nateghi, Noé F. Melo, Luca
Sangiuliano, Bert Pluymers and Wim Desmet
1
Dept. of Mechanical Engineering, KU Leuven, Belgium, claus.claeys@kuleuven.be
Abstract. Periodic structures, such as honeycomb core panels, combine excellent mechanical
properties with a low mass, making them attractive for application in transport and machine
design. However, the high stiffness to mass ratio of these lightweight panels may result in
unsatisfactory dynamic behaviour in that it may impair the panels' ability to reduce noise and
vibration levels.
Liu et al. demonstrated that inclusions of high density spheres with a rubber coating in a matrix
material result in low frequency sound isolation breaking the mass law [1]. These locally
resonant metamaterials require a high density of local resonators throughout the matrix
material, either spread randomly or periodically.
This chapter describes a novel metamaterial concept based on sandwich structures with
embedded resonant structures that exhibit vibro−acoustic stopband behaviour. A metamaterial
encapsulation is designed and produced to showcase the potential of this type of metamaterials
to improve acoustic insulation performance in targeted frequency ranges. As compared to
classic encapsulations made from homogeneous panels, the acoustic insertion loss from the
metamaterial is strongly increased within the stop band region of the metamaterial without
increasing the mass and hardly increasing the volume
4A.1 Introduction
Increasing customer expectations and more restrictive legal requirements turn the
acoustical behaviour of products into an important design criterion in the machine and
transportation industry as well as in the construction and consumer goods sector.
Ecological trends and the associated run for efficiency, however, increase the
importance of lightweight design and reduce the applicability of classical (heavy)
solutions to improve acoustic behaviour. In view of this challenging and often
conflicting task of merging acoustical and lightweight requirements novel acoustic
solutions are required. Ideally these novel solutions are easy to design and are
characterised by a low mass and compact volume along with a high reliability at an
affordable cost.
Vibro-acoustic metamaterials come to the fore as possible candidates for lightweight
material systems with superior noise and vibration insulation, be it at least in some
targeted and tuneable frequency ranges, referred to as stop bands. Contrary to
photonic crystals, stop bands in metamaterials don't rely on periodicity or Bragg
scattering and work on spatial scales much smaller than the wavelength [2]. The stop
bands induced in metamaterials result from resonant cells arranged on a
______________________________________________________________________________________
218
subwavelength scale and can be described based on the Fano-type interference
between incoming waves and the waves re-radiated by the resonant cells [3, 4].
Previous papers of the authors explain the working principles of stop bands based on
resonant metamaterials and list the driving parameters for stop band design [5].
In this chapter an innovative metamaterial concept to improve acoustic behaviour is

proposed; resonant structures are added to cavities of a periodic core sandwich
structure to create stop band behaviour. The next section describes the rationale
behind the proposed metamaterial concept, followed by a section describing the
numerical prediction of the stop band behaviour of this concept. The following
sections elaborate on the demonstrator design, measurements on this demonstrator
and the validation with numerical simulations. The chapter ends with the conclusions.
4A.2 Metamaterial creation by inclusion of resonant structures
Metamaterials with stopband behaviour are obtained through the inclusion of resonant
cells on a scale smaller than the structural wavelengths to be influenced [5]. Stop
band behaviour can thus be achieved through the introduction of any system that
introduces local resonant behaviour.
In view of applications, the goal is to find resonant systems which do not jeopardise
other requirements; structural integrity, light weight, use in contaminated
environment, fire-resistance... The kinds of resonant systems which are eligible
heavily depend on the structure to which the resonant systems have to be added.
Inspiration for a high potential structure to introduce stop band behaviour is sought in
the class of periodic lightweight structures, such as honeycomb core sandwich panels.
They are becoming attractive for application in transport and machine design due to
the combination of excellent mechanical properties with a low mass.
Fig. 1-A shows examples of sandwich structures; a core acts as spacer to create
distance between the skins such that a light structure with excellent stiffness
properties in bending is obtained. The core has as main role to create distance
between the skins as well as to resist forces perpendicular to the structure while the
skin is designed to show a high in plane strength; given the different requirements for
both, often the skin is made of a different material as the core. Different core layouts
are possible and two typical layouts are hexagonal or rectangular cores.
Fig. 1-A: Examples of sandwich structures; left a hexagonal core, right a rectangular core [6]
______________________________________________________________________________________
219
The internal cavities of these periodic core sandwich panels are an ideal host structure
to include resonant systems. In such configuration the resonant systems can be freely
designed for optimal acoustic behaviour in a targeted frequency zone while
safeguarding the excellent mechanical properties of the sandwich panels. The only
requirement for the resonant structures is the presence of a resonance mode; a huge
number of designs could be proposed, leaving room for optimisation towards e.g.
minimal weight, maximal attenuation or broadband attenuation.
For this investigation a straight forward design is chosen that resembles a mass-spring
system; two thin legs are used to connect a heavy mass to a host structure (Fig. 2-A).
The connection legs will determine the stiffness while the thick part of the resonator
will determine the mass of this resonant structure.
Fig. 2-A: Resonant structure used in this project to introduce stop band behaviour
Adding the resonant structures to the cavities of a periodic sandwich core (Fig. 3-A)
will introduce a stop band, and thus improved acoustic behaviour, for frequencies in
the vicinity of the resonance frequency of the resonant structure. Further analyses of
how the resonant structures introduce stop band behaviour, is the topic of next
section.
Fig. 3-A: Rectangular core without (left) and with (right) resonant structures
4A.3 Stop band prediction
From literature it is known that wave propagation through infinite periodic structures
can be investigated through unit cell modelling [7] [8]. Based on an undamped Finite
Element (FE) model of the unit cell and the application of periodicity boundary
conditions, dispersion curves for freely propagating waves in an infinite periodic
______________________________________________________________________________________
220
structure can be derived. Frequency zones, for which no solutions are found,
correspond to frequency zones without free wave propagation and thus a stop band
region.
Fig. 4-A: Dimensions in millimeter of the unit cell of the enclosure. The resonator leg thickness
(L) equals 1 mm, the resonator mass (M) equals 4 mm
Fig. 4-A shows the geometry of the unit cell that will be used to build a demonstrator.
The resonant structure makes up 30\% of the weight of the unit cell. Fig. 5-A depicts
the FE model of this unit cell. Linear Quad4 elements are used to represent the
resonant structure, core and skin. The resonator mass has a larger thickness than the
connection legs; different properties are assigned to both element groups. The
resonator mass element group is depicted in grey on the right side of Fig. 5-A.
Fig. 5-A: FE model of the unit cell of the demonstrator. The grey elements in the top view
(right) represent the resonator mass
Table 1-A: Material characteristics of the material of the unit cell in the numerical simulations
Name Symbol Value

Young’s modulus E 1.65 Mpa
Density ρ 950 kg/m3
Poisson’s ratio ν 0.4
The material characteristics of the acoustic enclosures are given in Table 1-A. The
enclosures are made through Selective Laser Sintering (SLS) of Polyamide, the
material and production characteristics are discussed in more detail in next section.
______________________________________________________________________________________
221
Since the main goal of the simulations is to get an indication whether stop band
behaviour is present rather than obtaining a detailed model, isotropy and linear
behaviour of the material are assumed.
The resonant structure is designed to have a pronounced low frequent bending mode
followed by subsequent modes higher in frequency. Fig. 6-A shows the first two
modes of the free unit cell which can be related to modes of the resonant structure.
Left the bending mode is shown which occurs at a frequency of 1272 Hz, the next
mode, shown at the right, occurs at 5052 Hz; this large shift in frequency between the
modes allows to correlate stop band behaviour with a certain resonance mode.
Convergence of the model was validated against a refined unit cell model of 6766
nodes and 6576 elements. The used model with 198 nodes and 168 elements has an
accuracy of 1 % on the first mode and 1.5 % on the second mode.
Fig. 6-A: Undeformed mesh (light) and the deformation of the first modes (dark) of the
resonant structures. Left: bending mode at 1272 Hz. Right: torsional mode at 5052 Hz
The unit cell allows derivation of the dispersion curves of the demonstrator. Fig. 7-A
shows the bending wave dispersion curves for the unit cell of the demonstrator in
comparison to the dispersion curves of the host structure without resonant structures.
The dispersion curves are similar except for a region around the resonance frequency
of the resonant system for which no dispersion curves exists; a stop band opens up
from 1065 Hz up to 1226 Hz. The shift with respect to the resonance frequency of the
resonant structure (1272 Hz), is explained by the different boundary conditions of the
unit cell during calculation of the resonance frequency.
Fig. 7-A: Dispersion curves of the demonstrator (—) in comparison with the dispersion curves
of the host structure (—)
______________________________________________________________________________________
222
4A.4 Demonstrator Example
To prove the potential of the introduced metamaterials concept to reduce acoustic

transmission, an acoustic enclosure making use of resonant structures is designed. A
variety of demonstrators can be thought of to determine the acoustic transmission
characteristics of a material. In the choice for a suitable demonstrator, some
determining factors are: unambiguous proof of concept, quantifiable effect and
engineering relevance. To obtain an unambiguous proof of concept it is crucial to
have as less unknown effects in the design as possible. Connections between different
components by bolts or combining parts together by glueing, is a likely source of
uncertainty; it is preferred to have a demonstrator built in one part such that these
unknowns are reduced. It was chosen to design acoustic enclosures as boxes with one
open side which can be placed over a small speaker. The acoustic transmission loss is
then determined by comparing sound radiation with and without enclosure.
Producing acoustic enclosures in one part is hard to achieve with traditional
manufacturing processes. Additive manufacturing is a production process which
allows producing complex parts without the need of an expensive mould, making it a
suitable production process for prototype design. Within the range of additive
manufacturing processes, Selective Laser Sintering (SLS) is chosen to create the
demonstrator. SLS is an additive manufacturing method where small powder particles
are melted together by means of a laser. The melting of adjacent material particles is
called sintering and by controlling the laser, a pattern can be sintered in a layer of
material powder. Unlike other additive manufacturing methods, such as fused
deposition modelling and stereolithography, the part being built is surrounded by
unsintered powder at all times. This unsintered powder acts as a support for the next
layers and facilitates building complex parts with jumps in geometry across the height
of the part. This possibility of building complex parts is the main reason why SLS
was chosen as production process.
Fig. 8-A: Demonstrator design; side (left) and bottom (right) view
______________________________________________________________________________________
223
Fig. 8-A shows a side and bottom view of the demonstrator design. One side of the
enclosure contains 8 x 8 unit cells. It should be noted that the corners of the
demonstrator are hollow and no resonators are added. The size of the enclosure is a
balance between obtaining a light demonstrator and still being relevant; the enclosure
is designed such that the inner dimensions make up a cube of 100 x 100 x 100 mm.
Fig. 9-A shows a produced version of this demonstrator.
Fig. 9-A: Pictures of the demonstrator
4A.5 Measurement results
The test set-up consists of a small loudspeaker placed on a wooden plate (Fig. 10-A).
The wooden plate is placed on an iron support in the centre of a semi-anechoic
chamber. Between the wooden plate and the loudspeaker, a small trim-like piece of
fabric is placed; the trim covers the cable of the loudspeaker which runs in a small
split in the plate.
Fig. 10-A: Picture of the test set up (left) and an intensity measurement of an acoustic
metamaterial enclosure on the test set up (right)
______________________________________________________________________________________
224
The sound power is evaluated on the surface of a surrounding box of 250 x 250 x 190
mm centred on the loudspeaker; the black dots on the wooden plate (Fig. 10-A)
indicate the width and length of the surrounding box. The acoustic power is evaluated
based on 5 intensity measurements, one on each side of the box, with a scanning
intensity probe.
The insertion loss is defined as:
𝑃!"#!!"#
𝐼𝑛𝑠𝑒𝑟𝑡𝑖𝑜𝑛 𝐿𝑜𝑠𝑠 = 10 log!" Eq. 1 − B
𝑃!"#!
where P_without is the radiated acoustic power radiated through the imaginary
evaluation box when the enclosure is not in place and P_with is the radiated acoustic
power when the enclosure is in place. The use of insertion loss as metric has as added
value that the results are not dependent on source spectrum.
Fig. 11-A compares the measured insertion loss of the metamaterial demonstrator to
the measured insertion loss of a regular enclosure with the same mass but build from
flat panels of 3.5 mm thickness. Between 700 and 1000 Hz the demonstrator
enclosure clearly outperforms the regular enclosure. The frequency zone of improved
acoustic behaviour is a bit lower in frequency than the stop band predicted by the
design of the resonant structures.
Fig. 11-A: Comparison between the measured insertion loss for the metamaterial demonstrator
and a demonstrator with flat side panels and an equivalent weight to the metamaterial
demonstrator.
Further investigation of the produced demonstrator explains the shift in frequency;

through optical measurements of dedicated test samples of the demonstrator unit cell
on a CNC Mitutoyo QuickVision Pro - 202 machine it was shown that the average
geometry is smaller than the nominal geometry [6][9]. This shift in geometry should
be taken into account in the numerical model; the shift from the nominal 1 and 4 mm
to the measured 0.83 and 3.85 mm thickness for the connecting leg and resonator
mass respectively leads to a shift in stop band frequencies from 1065 – 1226 Hz to
861 – 986 Hz, corresponding well with the measures zone of increased acoustic
insertion loss.
______________________________________________________________________________________
225
4B.6 Comparison with numerical models of the finite enclosure
To validate the measurement results a numerical counterpart is built for the test set-up
and the acoustic enclosure.
4B.6.1 Model of the test set-up
The wooden plate is modelled as a baffle: an infinite plane which is acoustically

perfectly reflecting. The model is shown in Fig. 12-A. This approach is justified since
the acoustic power is measured above the wooden plate and the set-up is positioned
within a semi-anechoic chamber. An impedance layer in the plane of the baffle
reflects the piece of trim on the wooden plate; a real impedance of 1.578 104 kg/m2
corresponding to a sound absorbing coefficient of 0.1 is added [10]. The piece of trim
has a size similar to the size of the acoustic enclosure. The impedance layer in the
numerical model covers the part of the baffle between the model of the loudspeaker
and the enclosure.
The loudspeaker is modelled as a rigid perfectly reflecting box with the same
dimension as the loudspeaker: 73 x 73 x 55 mm. The speaker mesh and the
impedance boundary have a maximal element length of 10 mm. Following the 6
elements per wavelength rule, this mesh is acoustically valid up to above 5000 Hz.
To model the sound source the a point source is used. The point source emits a
constant intensity across the frequency spectrum and is placed 5 mm above the centre
of the upper plane of the speaker. The upper surface of the speaker mesh is refined
since the point source is placed close to this surface. As explained in last section,
insertion loss is used as acoustic characteristic for the enclosures, which should be
independent of source spectrum.
Fig. 12-A: Left: model of the test set-up. Right: Mesh on a hemisphere with a 0.3 m radius
around the model of the loudspeaker to evaluate the radiated acoustic power in the numerical
model.
The numerically predicted radiated acoustic power is calculated as the sum of the
radiated acoustic intensity on a hemisphere with a 0.3 m radius which is centred over
the loudspeaker. Fig. 12-A, depicts the mesh, consisting of 105 elements and 113
nodes with a largest element edge equal to 0.1m. This mesh was validated by
comparison with refined meshes for the evaluation of the acoustic radiated power.
______________________________________________________________________________________
226
4B.6.2 Model of the acoustic enclosure
To compute the acoustic radiated power through the enclosure under excitation of the
point source within the enclosure a simulation is required in which an acoustic model
of the test set-up is coupled to a vibro-acoustic model of the enclosure. To reduce
computational cost, a modal-based vibro-acoustic coupling is applied; the structural
modes of the enclosure are derived for a detailed structural FE mesh after which these
structural modes are projected on an acoustic Boundary Element (BE) mesh of the
enclosure such that a coupled vibro-acoustic simulation can be performed. These
calculations are performed in the LMS VL.Acoustics software environment coupled
to a Nastran structural solver and a Sysnoise acoustical solver.
The FE model of the structure is derived through a repetition of the unit cell described
in section 3 of this article. Fig. 13-A shows the FE model of the acoustic enclosure
without outer skin. The total structural model with outer skin contains 63353 nodes
and 60720 linear Quad4 elements.
Fig. 13-A: Structural FE model of the acoustic enclosure with the outer skin not shown.
The element size of the structural mesh of the enclosure is driven by the smallest
geometry of the unit cell. This geometry is small compared to the wavelength in the
structure since the inclusion for metamaterial based stop bands needs to be smaller
than the governing wavelength in the structure.
The enclosure is supported by the wooden plate and the piece of trim. This is
modelled by removing the translational degrees of freedom of all elements in the
bottom plane of the enclosure. The motion parallel to the wooden plate is influenced
by the boundary as well; this is neglected in the numerical model. To include the
effect of damping both in the material as through the boundary, 1% of modal damping
is applied to all the modes.
The minimal size of the acoustic mesh is driven by the wavelength of sound in air.
Since the simulations are below coincidence, the element size requirements on the
acoustic mesh are less stringent than for the structural mesh. The acoustic mesh is
______________________________________________________________________________________
227
built as a remeshing of the skin of the enclosure with 15 x15 elements on each side
such that the obtained element length is similar to the model of the loudspeaker. This
remeshing reduces the number of elements for the outer skin from 4140 for the
structural model to 1125 elements for the acoustic model. Fig. 14-A compares the
obtained acoustic mesh (right) with the structural mesh (left).
Fig. 14-A: The structural model of the enclosure with skin shown (left) and the acoustic model
(right).
To compute the acoustic radiated power through the enclosure a BE modal-based

vibro-acoustic coupling is applied. This coupling requires the mode shapes, computed
for the structural model, shown in Fig. 14-A and on the left of Fig. 14-A, to be
mapped onto the acoustic mesh. The equivalent displacement on a node of the
acoustic mesh is calculated based on the movement of the 20 closest nodes on the skin
of the structural mesh which are within a 5.5 mm radius of the node on the acoustic
mesh. The contributions of these nodes of the structural mesh to the node on the
acoustic mesh are weighted based on the distance between the nodes on structural and
acoustic mesh.
The mapping of the modes of a structural model with skin, core and resonant
structures on an acoustic mesh without core and resonant structures means that the
resonant structures cannot be influenced by the acoustic medium in the simulation and
the other way around, that the movement of the resonant structures will not influence
the acoustic medium. This simplification was validated by comparing the obtained
results with this strategy to a simulation using the full structural mesh (skin, core and
resonant structure) as acoustic mesh: every node of the structural model coincided
with a node on the acoustic model. The difference showed to be negligible, indicating
that the mesh mapping is allowed and that the acoustic mesh size is fine enough.
This validation of the mesh mapping indicates that the surrounding air does not
influence the motion of the resonators in the numerical model. In reality however, the
relative motion of the resonant structure with respect to the host structure could lead
to added damping by air being squeezed through the gap between resonant structure
and core. The acoustic model applied does not account for possible viscosity effects.
______________________________________________________________________________________
228
4B.6.3 Validation results
A numerical simulation of the demonstrator with this geometry of the resonator

updated according to the measurements shows a good predicted insertion loss (Fig.
15-A). Details on the numerical model used can be found in [6] [11]. The depth of the
stop band is overestimated in the simulations, but this might be related to a wrong
estimation of the amount of damping in the material as well as due small variation in
the resonance frequencies of the resonant structures; both the damping as a spread on
the resonance frequencies can have a strong effect on strength of acoustic reduction in
the stop zone [5] [6].
Fig. 15-A: Comparison between measured and numerically predicted insertion loss of the
metamaterial demonstrator. The geometry of the resonant structures was updated to comply
with measurements results on the geometry of a unit cell.
4B.7 Conclusions
This section introduces a novel method to create resonant metamaterials for acoustic
insulation; the inclusion of resonant structures within the core of a sandwich structure.
Through the numerical and experimental validation of a demonstrator it is shown that
this results in a frequency zone of increased acoustic insertion loss with respect to
equivalent materials of the same weight. Unit cell modelling allows a quick
estimation of the location of the stop band frequencies and can be used as a tool to
assess changes in resonant structure design. This metamaterial concept allows the
combination of light weight, compact mass and good acoustic behaviour along with
other technological benefits such as integration in structural parts, use in harsh
environments and easy designable beneficial frequency ranges. To see and hear this
potential in a movie, the interested reader is referred to a movie which can be seen on
the following link (http://youtu.be/tOch_GsGaXg).
______________________________________________________________________________________
229
Acknowledgements
The authors would like to acknowledge the European Commission for their support
through the ENLIGHT-project (http://www.project-enlight.eu/) and the EMVeM-
project (http://www.emvem.org/) and the KU Leuven Research Fund for their support
through the IOF. E. Deckers and L. Van Belle are respectively a postdoctoral and
doctoral fellow of the Fund for Scientific Research Flanders (F.W.O.). This Fund is
also gratefully acknowledged for its support in the G.0371.12 research project. The
research of N. F. Melo is funded by a PhD Scholarship Sciences without Borders
CNPq Brazil (201414/2014-7).
4B.8 References
[1] Z. Liu, X. Zhang, Y. Mao, YY Zhu, Z. Yang, CT Chan, and P. Sheng, Locally
resonant sonic materials, Science 289 (2000), no. 5485, 1734–1736.
[2] J B. Pendry, D.J. Holden, A J.and Robbins, and W.J. Stewart, Magnetism from
conductors and enhanced nonlinear phenomena, Microwave Theory and
Techniques, IEEE Transactions on 47 (1999), no. 11, 2075–2084.
[3] C. Goffaux, J. Sánchez-Dehesa, A.L. Yeyati, P. Lambin, A. Khelif, JO Vasseur,
and B. Djafari-Rouhani, Evidence of fano-like interference phenomena in locally
resonant materials, Physical Review Letters 88 (2002), no. 22, 225502.
[4] F. Lemoult, N. Kaina, M. Fink, and G. Lerosey, Wave propagation control at the
deep subwavelength scale in metamaterials, Nature Physics 9 (2012), 55–60.
[5] C. Claeys, K. Vergote, P. Sas, and W. Desmet, On the potential of tuned
resonators to obtain low frequency vibrational stop bands in periodic panels,
Journal of Sound and Vibration 332 (2013), no. 6, 1418–1436.
[6] C. Claeys, Design and analysis of resonant metamaterials for acoustic insulation,
PhD Thesis, KU Leuven, 2014.
[7] L. Brillouin, Wave propagation in periodic structures, 2nd ed., McGraw-Hill Book
Company, 1946.
[8] R.S. Langley, A note on the force boundary conditions for two-dimensional
periodic structures with corner freedoms, Journal of Sound and Vibration 167
(1993), 377–381.
[9] C. Claeys, B. Pluymers, P. Sas, W. Desmet. Design of a resonant metamaterial
based acoustic enclosure. ISMA2014 - USD2014. Leuven, Belgium, 15-17
September 2014 (art.nr. 561) (pp. 3351-3358).
[10] Vivolo, M., Vibro-acoustic characterization of lightweight panels by using a
small cabin, Ph.D. thesis, KU Leuven, Departement Werktuigkunde, 2013.
[11] C. Claeys, E. Deckers, B. Pluymers, and W. Desmet, A lightweight vibro-
acoustic metamaterial demonstrator: Numerical and experimental investigation,
Mechanical Systems and Signal (2016), 70-71, 904-918.
______________________________________________________________________________________
230
Annex 4B
Motion equations of mechanical systems used in common
engineering applications
Sorin VLASE1
1
Dept. of Mechanical Engineering, Transilvania University of Brasov, Romania
svlase@unitbv.ro
4B.1 Motion equations of an one-dimensional finite element
4B.1.1 General considerations
In what follows the motion equations governing the evolution in time of a truss finite
element, participating along with the body to the general rigid body motion of the
system are being determined [1-7]. In the presentations important parts represent the
author’s contribution to the development of this field [8-19]. Different applications
are presented in [20-29].
The method used will be the method of Lagrange equations. In order to be able to use
this method, the kinetic energy of the considered finite element, the internal energy
and the work of the distributed and concentrated loads (for obtaining the Lagrangian)
will be calculated.
An application of the method is presented in [30]. Experimental results are obtained
by [31] and [32]. Mathematical methods to solve the obtained equations are presented
in [33] and [34].
The independent coordinates will be the vectors of nodal displacements which may
include the displacements in the three directions of bar ends and derivatives of these
values having the significance of rotations and curvatures. The number of
independent coordinates chosen depends on the bar hypothesis used; they will
determine the values that count for the model description and for the way of choosing
the interpolation functions. If a bar is flexible only on its own axis and rigid on the
other two axes only the axial displacements of bar ends will be considered as being
independent coordinates. But if the bar is rigid in axial direction and flexible in the
other directions, the displacements in the directions perpendicular on the bar as well
as the rotations of the bar cross - section in two directions perpendicular on the bar are
being considered.
______________________________________________________________________________________
231
Fig. 1-B. One-dimensional finite element with three-dimensional motion
The aim of the paper is to determine the general motion equations and not to make all
the particularizations determined by the idealizations and models chosen. For a type
of finite element often used in engineering we shall also obtain the form specific for
this case of motion equations
4B.1.2 Lagrangian of the Mechanical System

Within the multibody system containing the finite element the latter has a general
three-dimensional rigid body motion. It is assumed that the rigid motion of the
mechanical system was previously determined considering all the rigid elements of
the multibody system. We start from the hypothesis that the elastic deformations are
small and would not influence the general motion of the system of bodies. Thus it is
assumed that the velocities and accelerations field for each element of the mechanical
system are known.
Let’s consider a truss finite element, having at both ends the nodes i and j. δi and δ j
represent the displacements of bar ends (Fig.1-B). The finite element is related to a
local reference system Oxyz participating to the rigid motion of the bar. We know the
velocity and acceleration of the origin of the mobile reference system with respect to
the fixed reference system Ox1 y1 z1. The angular velocity and angular acceleration of
the mobile reference system is also considered as being known.
The displacement of a point M of bar δ (u,v,w) can be expressed in terms of nodal
displacements as follows:
______________________________________________________________________________________
232
⎧u ⎫
⎪ ⎪ ⎧δ ⎫
δ = ⎨ v ⎬ = N δe = N ⎨ 1 ⎬ Eq. 1-B
⎪w⎪ ⎩δ2 ⎭
⎩ ⎭
where the vector of nodal displacements δe is:
⎧δ ⎫
δe = ⎨ 1 ⎬ Eq. 2-B
⎩δ2 ⎭
and the matrix N contain shape functions. The lines of the matrix N corresponding
to the displacements u, v and w are named N (u ) , N (v ) and N ( w ) :
⎡ N (u ) ⎤
⎢ ⎥
N = ⎢ N (v ) ⎥ Eq. 3-B
⎢ N (w ) ⎥
⎣ ⎦
The vectors of displacements of the nodes at bar ends (right and left bar ends) have
been named δ1 and δ2 . Thus we have:
⎧u ⎫ ⎡ N (u ) ⎤
⎪ ⎪ ⎢ ⎥⎧δ ⎫
δ = ⎨ v ⎬ = N δ e = ⎢ N (v ) ⎥ ⎨ 1 ⎬ Eq. 4-B
δ
⎪ w⎪
⎩ ⎭ ⎢ N (w ) ⎥ ⎩ 2 ⎭
⎣ ⎦
If v and w are the displacements of a bar point in the directions Oy and Oz

respectively (displacements perpendicular on the bar axis) we shall have the equations
known from the mechanics of continuous media ([35][36][37]):
dw dv
β =− ;γ = Eq. 5-B
dx dx
Considering the shape functions, the rotations of the bar ends can be expressed as
follows:
______________________________________________________________________________________
233
d
β =− (N w δe ) = − N w' δe ; γ = d (N v δe ) = N v' δe ; Eq. 6-B
dx dx
In what follows index G applied to a vector or a matrix means that the respective
value is expressed in the global reference system and index L means that the
respective value is expressed in the local reference system.
As a result of the displacement the position vector of point M becomes in the
deformed position M' and can be expressed by:
⎧u ⎫ ⎧x + u⎫
⎪ ⎪ ⎪ ⎪
rM ', L = rM , L + δ = rM , L + ⎨ v ⎬ = ro, L + ⎨ v ⎬ Eq. 7-B
⎪w⎪ ⎪ w ⎪
⎩ ⎭ ⎩ ⎭
or, with respect to the global reference system:
⎧u ⎫ ⎧ x⎫ ⎧u ⎫ ⎧ x⎫
⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪
rM ',G = rM ,G + R⎨ v ⎬ = ro,G + R⎨0⎬ + R⎨ v ⎬ = ro,G + R⎨0⎬ + R N δe, L Eq. 8-B
⎪w⎪ ⎪0⎪ ⎪w⎪ ⎪0⎪
⎩ ⎭ ⎩ ⎭ ⎩ ⎭ ⎩ ⎭
where the matrix R expresses the transition of the mobile reference system Oxyz to
the fixed reference system O'XYZ . The velocity is given by the relation:
⎧x⎫
⎪ ⎪
v M ,G = r!M ',G = r!o,G + R⎨0⎬ + R! N δe, L + R N δ!e, L
! Eq. 9-B
⎪0⎪
⎩ ⎭
The kinetic energy expression for the whole bar is:
L
1
Ec =
20∫ ( )
ρ Ar!MT ',G r!M ',G + ω'TL I ω' L dx =
______________________________________________________________________________________
234
L
⎛ ⎧ x⎫ ⎧x⎫ ⎞
1 ⎜ T T !⎪ ⎪ ⎪ ⎪⎟
= ∫ ⎜ r!o,G r!o,G + 2r!o,G R⎨0⎬ + 2r!o,G R Nδe + 2r!o,G R N δ!e + [x 0 0]R! R! ⎨0⎬ ⎟ρAdx +
T ! T T
2 0⎜ ⎪0⎪ ⎪0⎪ ⎟
⎝ ⎩ ⎭ ⎩ ⎭⎠
L
1
(
+ ∫ [x 0 0]R! T R! N δe + [x 0 0]R! T R N δ!e + δe N T R! T R! N δe ρAdx +
20
T
)
L L
1 1
+
20∫ (T T
) 20
(
2δe N T R! T R N δ!e + δ!e N T RT R N δ!e ρAdx + ∫ ω'TL I ω' L ρAdx )
Eq. 10-B
where:
⎡ I xx 0 0 ⎤
⎢ ⎥
I =⎢ 0 I yy − I yz ⎥ Eq. 11-B
⎢0 − I yz I zz ⎥⎦
⎣
I yy and I zz represent moments of inertia of the bar cross section about co-ordinate
axis Oy and Oz respectively of a reference system with its origin in the mass centre of
the element dm = ρAdx ( ρ - density); I yz is the centrifugal moment of inertia and
I xx is the inertia moment about the co-ordinate axis Ox. Since we have chosen y and
z as principal directions of inertia I yz = 0 , we have:
⎡I x 0 0⎤
I = ⎢⎢ 0 Iy 0 ⎥⎥ Eq. 12-B
⎢⎣ 0 0 I z ⎥⎦
where, for the sake of simplicity I xx = I x , I yy = I y , I zz = I z . The angular velocity

of the element dm is:
______________________________________________________________________________________
235
⎧ω1L ⎫ ⎧α! ⎫ ⎧ω1L ⎫
⎪ ⎪ ⎪ ⎪ ⎪ ⎪
ω' L = ⎨ω2 L ⎬ + ⎨β! ⎬ = ⎨ω2 L ⎬ + N *δe Eq. 13-B
⎪ω ⎪ ⎪ γ! ⎪ ⎪ω ⎪
⎩ 3L ⎭ ⎩ ⎭ ⎩ 3L ⎭
where: ω1L , ω 2 L , ω3L are the components of the angular velocity vector related to
the mobile reference system.
⎧α ⎫ ⎧α! ⎫
⎪ ⎪ ⎪ ⎪
*
⎨β ⎬ = N δe ; ⎨β! ⎬ = N *δ!e Eq. 14-B
⎪γ ⎪ ⎪ γ! ⎪
⎩ ⎭ ⎩ ⎭
⎡ N (*α ) ⎤
* ⎢ * ⎥
where: N = ⎢ N (β ) ⎥ . Easily can be noted that: N (*β ) = N w' and N (*γ ) = N v' .
⎢ N (*γ ) ⎥
⎣ ⎦
In the written relations we can notice the occurrence of derivatives of the matrix for
positional relation to the global reference system. These derivatives represent angular
velocities and accelerations. For a better understanding we shall express these
derivatives according to [15].
Next the internal energy stored in the bar as a result of deformation shall be calculated
[38][39][40] and [41].
The energy due to bending is given by the expression:
2 2
1 ⎡ ⎛ d 2v ⎞ ⎤
L L
⎛ d 2w ⎞ 1
20⎢
⎢ ⎜ ⎟ ⎜ ⎟ ⎥
20
[
E pi = ∫ EI y ⎜ 2 ⎟ + EI z ⎜ 2 ⎟ dx = ∫ EI y β ' 2 + EI z γ ' 2 dx ] Eq. 15-B
⎣ ⎝ dx ⎠ ⎝ dx ⎠ ⎥⎦
If we write:
w = N (w )δe ; v = N (v )δe Eq. 16-B
then:
L
1⎡ ⎛ T T ⎤
E pi = ⎢∫ ⎜ EI y N (''w) N (''w) + EI z N (''v ) N (''v ) ⎞⎟dx ⎥ δe Eq. 17-B
2 ⎣⎢ 0 ⎝ ⎠ ⎥
⎦
______________________________________________________________________________________
236
The energy due to axial deformation is:
L 2 L
1 ⎛ du ⎞ 1 T
E pa
2 0 ⎝ dx ⎠ 2
(
= ∫ EA⎜ ⎟ dx = δe ∫ N u 'T N u ' EAdx δe ) Eq. 18-B
0
The deformation energy due to torsion is:
L 2 L
1 ⎛ dα ⎞ 1 T⎛ * T *
⎞
E pt = ∫ GI x ⎜ ⎟ dx = δe ⎜⎜ ∫ N (α ) ' N (α ) ' GI x dx ⎟⎟ δe Eq. 19-B
20 ⎝ dx ⎠ 2 ⎝0 ⎠
The axial load P in an axial section of the bar gives the energy if in a first
approximation the axial deformations are neglected:
1
L ⎡⎛ dv ⎞ 2 ⎛ dw ⎞ 2 ⎤
Ea = P ⎟ ⎥dx Eq. 20-B
2 ∫0
tot ⎢⎜ ⎟ +⎜
⎣⎢⎝ dx ⎠ ⎝ dx ⎠ ⎦⎥
where Ptot represents the axial force in the bar cross section at distance x. The force
components acting at the right bar end considered in the local coordinate system are
represented by Px, Py=0, Pz=0 Beside these components, the value of P and the
components of the inertia forces acting upon the portion of the bar between x and L
are being determined. In order to calculate them we determine the acceleration of the
current point of the bar in the rigid motion (Fig.2-B):
⎧ x⎫ ⎧ x⎫ ⎧ x⎫ ⎧ x⎫
⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ Eq. 21-B
a G = a o ,G + ε G R⎨0⎬ + ωG ωG R⎨0⎬ = a o,G + R ε L ⎨0⎬ + RωL ωL ⎨0⎬
⎪0⎪ ⎪0⎪ ⎪0⎪ ⎪0⎪
⎩ ⎭ ⎩ ⎭ ⎩ ⎭ ⎩ ⎭
The inertia force is given by:
L L L ⎧s ⎫ L ⎧s ⎫
⎪ ⎪ ⎪ ⎪
Fi = − ∫ aG dm = − ∫ ao,G ρAds − ∫ R ε L ⎨0⎬ρAds − ∫ RωL ωL ⎨0⎬ρAds =
x x x ⎪0⎪ x ⎪0⎪
⎩ ⎭ ⎩ ⎭
______________________________________________________________________________________
237
⎧ X!!o ⎫ ⎧ L2 − x 2 ⎫ ⎧ L2 − x 2 ⎫
⎪ !! ⎪ 1 ⎪ ⎪ 1 ⎪ ⎪
= − ⎨ Yo ⎬ ρA(L − x ) − Rε L ⎨ 0 ⎬ ρA − RωL ωL ⎨ 0 ⎬ ρA Eq. 22-B
⎪ Z!! ⎪ 2 ⎪ 0 ⎪ 2 ⎪ 0 ⎪
⎩ o⎭ ⎩ ⎭ ⎩ ⎭
Fig. 2-B. Determination of the axial inertia force
⎧µ x ⎫ ⎧ X!!o ⎫ ⎧1⎫ ⎧1⎫

⎪ !! ⎪ 1 1
{µ} = ⎨µ y ⎬ = − ρAL ⎨ Yo ⎬ − Rε L ⎨0⎬ρAL − RωL ωL ⎪⎨0⎪⎬ρAL2
⎪ ⎪ ⎪ ⎪ 2
Eq. 23a-B
⎪µ ⎪ ⎪ Z!! ⎪ 2 ⎪0⎪ 2 ⎪0⎪
⎩ z⎭ ⎩ ⎭ o ⎩ ⎭ ⎩ ⎭
⎧λ x ⎫ ⎧ X!! o ⎫
{λ} = ⎨λ y ⎬ = ρA⎪⎨ Y!!o ⎪⎬
⎪ ⎪
Eq. 23b-B
⎪λ ⎪ ⎪ Z!! ⎪
⎩ z⎭ ⎩ o⎭
⎧ν x ⎫ ⎧1⎫ ⎧1⎫
⎪ ⎪ 1 1
{ν } = ⎨ν y ⎬ = Rε L ⎨0⎬ρA + RωL ωL ⎪⎨0⎪⎬ρA
⎪ ⎪ Eq. 23c-B
⎪ν ⎪ 2 ⎪0⎪ 2 ⎪0⎪
⎩ z⎭ ⎩ ⎭ ⎩ ⎭
We obtain the internal energy due to inertia in the form of:
L
1 T⎡ T T ⎤
2
( ⎝
)
Ea = δe ⎢∫ Px + µ x + λx x + ν x x 2 ⎛⎜ N (*γ ) N (*γ ) + N (*β ) N (*β ) ⎞⎟dx⎥ δe
⎠ ⎥
Eq. 24-B
⎣⎢ 0 ⎦
1 T G
Ea = δe k e δe Eq.25-B
2
______________________________________________________________________________________
238
where the notation for k eG is obvious. The total internal energy is:
T
(
E p = δe ki + k a + kt + k eG δe ) Eq. 26-B
The external work of distributed loads is:

L
(
W = ∫ p x u + p y v + p z w + m xα + m y β + m z γ dx = )
0
L
⎡N⎤
= ∫ px [ py pz mx my ]
mz ⎢ * ⎥ δe dx Eq. 27-B
0 ⎣N ⎦
and that of concentrated loads q e TL in the nodes is:
T
W c = q e L δe Eq. 28-B
The Lagrangian for the element is[42][43][44][45][46] and [47]:
L = Ec − E p − E a + W + W c . Eq. 29-B
4B.2 Two-dimensional finite element
4B.2.1 Motion’s equations for a two-dimensional finite element
Let's consider a two-dimensional finite element with a plan parallel motion. The type
of finite element which will be used will determine the shape functions and the final
form for the matrix coefficients. In what follows it is considered that the deformations
are small enough not to influence the general rigid motion of the system.
Both the problem of the rigid motion of the system and the field of velocities and
accelerations for each two-dimensional element of the multicorp systems are
considered to be solved ([34]). The finite element is related to the local coordinates
system Oxy, which is mobile and which participates to the whole motion (Fig. 3-B),
so it has a known rigid motion. Let's consider v o ( X! o ,Y!o ,0 ) being the velocity
______________________________________________________________________________________
239
and a o ( X!! o ,Y!!o ,0 ) being the acceleration of the origin of the mobile reference
system related to the fixed reference system OXY, to which the whole motion of the
mechanical system relates. We will consider ω (0, 0, ωz ) as being the angular
velocity of the solid containing the finite element and ε (0, 0, ε ) as being the
angular acceleration of the same solid.
Fig. 3-B. A two-dimensional finite element with plan motion
The change in the values of the position vector r M ,G of a random point M of the
finite element when going from the local Oxy reference system into the global fixed
OXY reference system is obtained using a rotation transformation matrix R. If we
consider r M ,G as being the position vector of the M point, we have:
r M ,G = r O , G + r G = r O , G + R ⋅ r L Eq. 30-B
where we consider G defining the vectorial entities which have their components
relating to the global reference system and L defining the vectorial entities which
have their components relating to the local reference system.
The rotation transformation matrix R has, in this situation, a very simple form:
______________________________________________________________________________________
240
⎡cosθ − sin θ ⎤
R =⎢ Eq. 31-B
⎣ sin θ cosθ ⎥⎦
If the M point is subjected to a displacement f L , transforming into the M' point, we

will have:
r M ',G
=r O ,G
+ R⋅(r L+ f L ) Eq. 32-B
where r M ',G is the position vector of the point M' with its components relating to the
global reference system. The continuous displacement field f (x,y) L is

approximated in the finite element method, depending on the nodal displacements,
using the relationship:
f L = N (x,y) δ e (t) L Eq. 33-B
where the elements of the N matrix (which contains the interpolation functions)
depend on the type of the chosen finite element. The velocity of the M' point, related
to the fixed coordinate system, will be:
v M ',G
= r! O + R! r L+
! f
R L+
R f! L=
r! O + R! r L+
! N δ
R e, L + R N δ! e, L
Eq. 34-B
The velocity components are defined related to the global coordinates system. The
equations of motion will be obtained in the local coordinates system.
The kinetic energy of the considered element will be determined using the
relationship:
1 1
Ec = ∫ ρv 2 dV = ∫ ρ v T
M ', L
v M ', L dV. Eq. 35-B
2 V 2 V
When we determined the kinetic energy, we could have been representing the velocity
in the global coordinates system, the result for the kinetic energy remaining the same.
______________________________________________________________________________________
241
This happens because the scalar product of two vectors remains the same, no matter
which reference system we use.
The deformation energy is:
1
Ep = σT ε dV . Eq. 36-B
2 ∫V
For the ease of understanding, we remember that the generalized Hooke law is:
σ = D·ε Eq. 37-B
where, for a two-dimensional finite element (plane state of deformation) we have:
⎡ ⎤
⎢1 − µ 0 0 ⎥
E ⎢ 0
D = 1− µ 0 ⎥ Eq. 38-B
(1 − µ )(1 − 2µ ) ⎢ 1 − 2µ ⎥⎥
⎢ 0 0
⎣ 2 ⎦
and the relation between the specific deformations and the finite deformations can be
expressed using the relationship:
ε=a· f Eq. 39-B
where a is a differentiation operator:
⎡∂ ⎤
⎢ 0 0⎥
⎢ ∂x ⎥
a ⎢ ∂ Eq. 40-B
= 0 0⎥
⎢ ∂y ⎥
⎢∂ ∂ ⎥
⎢ 0⎥
⎣⎢ ∂y ∂x ⎦⎥
Considering all this, for the deformation energy we have:
______________________________________________________________________________________
242
1 T
Ep = δ k e δ e, L dV Eq. 41-B
2 ∫V
e, L
with k e being the stiffness matrix:
T
k e = ∫ N a T D T a N dV. Eq. 42-B
V
If we consider p = p (x,y) as being the vector for the distributed forces, then the
external machine work of those forces is:
W =∫ p T
L f L dV=(
∫ p T
L N dV) δ e, L . Eq. 43-B
V V
The nodal forces q T

e give an external machine work:
W c=q T
e, L δ e, L . Eq. 44-B
The Lagrangian for the considered element is:
L = Ec − E p + W + W c =
1 1
ρ r! TO r! O p δ e, L + q
T T T
∫ δ k e δ e, L dV N δ e, L
2 ∫V
= dV - +( dV)
2 V
e, L ∫
V
L e, L
Eq. 45-B
The equations of motion are obtained by applying the Lagrange equations. After some
ordering, we can write the equations of motion for the finite element like this:
T
( N ∫ N ρdV) δ!! e, L + 2 ( ∫ N T RT R! N ρdV) δ! e, L +
V V
!! N ρdV)
+ ( k e + N T RT R δ e, L =
∫
V
______________________________________________________________________________________
243
=q e+ N p L dV – ( ∫ N ρdV) R !r! O – ∫ N T RT R
!! r ρdV
T T T
∫
V V V
Eq. 46-B
We can write the equations of motion in concentrated form:
m e δ!! e, L + 2 c e δ! e, L + [ k e + k e (ε) + k e (ω2)] δ e, L =
= q e + q *
e ,L – q i
e, L (ε) – q i
e, L (ω2) – m i
Oe
R T !r! O Eq. 47-B
where we considered:
T T
q *
e ,L =
∫ N p L dV ; m
i
Oe = N
∫
T
ρdV ; m e =
∫ N N ρdV = m 11 +m 22
V V V
c e (ω)= ∫ N T RT R! N ρdV ; k e (ε) + k e (ω2) = ∫N

T !! N ρdV
RT R
V V
q i
q i T !! r ρdV
RT R
e, L (ε) + e, L (ω) =
∫N
V
Eq. 48-B
The matrix products where the rotation transformation matrix appears can easily be
calculated, as this matrix is determined by only one element, the angle θ. The anti-
symmetric matrix:
! ⎡0 − ω⎤ ⎡0 − 1⎤
ω= R RT =⎢ = ω Eq. 49-B
⎣ω 0 ⎥⎦ ⎢1 0 ⎥
⎣ ⎦
represents the angular velocity operator corresponding to the angular velocity ω. We

also have:
!! R T = ε ⎡
0 − 1⎤ ⎡1 0⎤
R ⎢ ⎥ − ω2 ⎢ ⎥ Eq. 50-B
⎣1 0 ⎦ ⎣0 1 ⎦
In both the global and local reference systems, the angular velocity and angular
acceleration vectors have the same components.
______________________________________________________________________________________
244
If we consider N (1) and N ( 2) being the rows of the matrix N, with the notations:
m 1x =
T
∫ N (1) x ρdV ; m 1y = N
∫
T
(1) y ρdV ; m 11 = N
∫ (1)
T
N (1) ρdV
V V V
m 2x =
∫N
T
( 2) x ρdV ; m 2y =
∫N
T
( 2) y ρdV ; m 22 =
∫N ( 2)
T
N ( 2 ) ρdV
V V V
Eq. 51-B
we can obtain the equations of motion for the finite element, with explicit
dependencies to the angular velocity ω and angular acceleration ε. Previously, we
solve the integrals with the given notations and we can obtain a detailed form for the
equations of motion. Therefore, we have:
m e = (m 11
+m 22
) e; c e (ω) = ω (m21 − m12 ) e
k e (ε) + k e (ω2) = ε (m21 − m12 )e − ω 2 (m11 − m22 ) e
q i
e, L (ε) + q i
e, L (ω) (
= ε m 2 x − m1 y ) e - ω 2 (m1 x + m 2 y ) e Eq. 52-B
The equations of motion will be, in this case:
(m 11
+m 22
) e δ!! e, L + 2ω ( m 21
–m 12
) e δ! e, L +
+[ k e + ε m21 − m12
( ) e - ω2 (m11 + m22 ) e ] δ e,L =
(
= q eext + q eliaison – ε m 2 x − m1 y ) e + ω 2 (m1x + m2 y ) e – m i
Oe
R T !r! O Eq. 53-B
The unknowns in the finite element analysis of such a system are of two kinds: nodal
displacements and contact forces. Using a correct assembly of the equations of
motion, written by each finite element, the algebraic unknowns (the contact forces)
can be removed [13][14][46][48]. The equations of motion for the whole multibody
system will be possible to be written as a system of 2nd order non-linear differential
equations. The matrix coefficients of this system of equations have the following
properties:
• The inertial matrix m is symmetric;
• The damping matrix c is a skew symmetric matrix. The terms defined by
this matrix represent the Coriolis accelerations, , due to relative motions of
______________________________________________________________________________________
245
nodal displacements with respect to the mobile reference coordinate system,
linked to the moving parts of the system studied;
• The stiffness matrix k contains both symmetric and skew symmetric terms.
Moreover, this matrix can have singularities due to the rigid motion of the
system that have to be removed before conducting the study of the system.
4B.2.2 Modal analysis of a rotating disk
In the following there are presented results obtained by the author in the case of a
particular case, when we study a rotating homogeneous disk [30].
Let’s consider a rotating disk. If we consider a stationary motion, the disk will rotate
with an angular velocity ω. In a very small interval ∆t we may consider that the
angular velocity is constant and we intend to achieve the modal analysis of rotating
disk. We will use the method previously presented for carrying out the calculation and
compare the result with that obtained by applying the classic version of the method of
finite elements. We consider two cases:
Table 1-B. Eigenvalues of a rotating disk
Mode number
Classic model - Proposed model –
(without rigid body
eigenvalues (Hz) eigenvalues (Hz)
mode)
4 5514 5865
5 5514 5865
6 7045 7323
7 7045 7323
8 8959 9180
9 8959 9180
10 9416 9626
11 10641 10827
12 10641 10827
13 11713 11883
14 11713 11883
15 12767 12923
16 14284 14423
17 14284 14423
18 14910 15044
______________________________________________________________________________________
246
Fig. 4-B. Eigenmodes 4, 5 and 6
Fig. 6-B. Eigenmodes 10,11 and 12
______________________________________________________________________________________
247
A. The rotating disk has a plan motion and the finite element is in a plane
displacement field. The eigenvalues are presented in Table 1-B and the
eigenvector in Fig.4-B-Fig.8-B.
B. The rotating disk has a plan motion and the finite element is in a general
displacement field. The eigenmodes are presented in Fig.9-B - Fig.13-B.
______________________________________________________________________________________
248
The additional terms in the equations of motion will influence the dynamic response
of the system. It may happen that a resonant state or a loss of stability state to be
reached. The modifications of eigenvalues considering a rotating disk (along its own
axis) is presented in the paper. An upwards displacement for all the eigenvalues can
be observed - this happens because an increase in stiffness takes place in this situation
(rotation), due to the inertial forces. The vibration modes remain practically the same,
while a small change in the amplitudes can be observed. The presence of inertial and
______________________________________________________________________________________
249
Coriolis effects can significantly modify, in some situations, the dynamic response of
the system.
4B.3 References
[1] Thomson, B.S., Sung, C.K., A survey of Finite Element Techniques for
Mechanism Design. Mech.Mach.Theory, 21, nr. 4, p. 351–359 (1986).
[2] Nath, P.K., Ghosh, A., Kineto-Elastodynamic Analysis of Mechanisms by Finite
Element Method, Mech.Mach.Theory, 15, pp. 179 (1980a).
[3] Nath, P.K., Ghosh, A., Steady-state Response of Mechanisms with Elastic Links
by Finite Element Method. Mech.Mach.Theory, 15, pp. 199 (1980b).
[4] Nguyen, V.K., Zur Dynamischen Analyse Ebener Mechanismen mit mehrere
Freitsgraden unter Beruecksichtigung der Gestell Schwingungen Ebener
Mechanismen. St.Cerc.Mec.Apl., Tom 29,Nr.3 (1984).
[5] Nguyen, V.K., Ueber der Einfluss von Schwingungen des Gestells auf die
Dynamischen Stabilitaetsbedigungen und Periodischen Schwingungen Ebener
Mechanismen. St.Cerc.Mec.Apl., Tom 29,Nr.3 (1984).
[6] Pandrea, N., Le calcul des forces dans les mécanismes à barres linéairment
élastiques. Rev.Roum.Sci.Tech-Méc.Appl., Tom 33, Nr.1, p.73 (1988).
[7] Pandrea,N., Nicolaescu, N., Kineto-static Analysis of a Plane Mechanism with
Joints using Finite Element Method. (In Romanian). Mechanisms and Mechanical
Transmission, Timisoara, p.255 (1980).
[8] Vlase, S., On the Motion Equation of the Mechanism using Finite Element
Method (In Romanian). PRASIC Symposium, Brasov,vol.I, p.189 (1982).
[9] Vlase, S., Motion Equations for a Planar Finite Element (In Romanian). PRASIC
Symposium, Brasov, Vol.I, p.183 (1982).
[10] Vlase, S., Elastodynamische Analyse der Mechanischen Systeme durch die
Methode der Finiten Elemente. Bul. Univ. Brasov, p.1-6 (1984).
[11] Vlase, S., Ein Ebenes Finiten Element in Zustand einer Membrane fuer
Elastodynamische Analyse der Mechanischer Systeme. Bul. Univ. Brașov, p.547
(1985).
[12] Vlase, S., Bewegungsleichungen fuer ein Finit Element mit allgemeiner
Bewegung.Bul. Univ.Brașov, seria A, Vol.XXVIII, p.19-26 (1986).
[13] Vlase, S., A Method of Eliminating Lagrangean Multipliers from the
Equations of Motion of Interconnected Mechanical Systems. Journal of applied
Mechanics, ASME Transactions, vol. 54, nr. 1 (1987).
[14] Vlase, S., Elimination of Lagrangean Multipliers. Mechanics Research
Communications, vol. 14, pp. 17–22 (1987).
[15] Vlase, S., Contributions to the elastodynamic analysis of the mechanisms
with the finite element method (In Romanian). Ph.D., Transilvania University
(1989).
[16] Vlase, S., Finite Element Analysis of the Planar Mechanisms: Numerical
Aspects. In Applied Mechanics - 4. Elsevier, pp.90-100 (1992).
[17] Vlase, S., Teodorescu-Draghicescu, H., Calin, M.R., Scutaru, M.L.,
Advanced Polylite composite laminate material behavior to tensile stress on weft
______________________________________________________________________________________
250
direction. Journal of Optoelectronics and Advanced Materials, Vol. 14, No. 7- 8,
p. 658 – 663 (2012).
[18] Vlase, S., Teodorescu,P.P., Elasto-dynamics of a solid with a general “rigid”
motion using FEM model. Part I.Rom. Journ. Phys., Vol. 58, Nos. 7–8, P. 872-
881, Bucharest, (2013).
[19] Vlase, S., Teodorescu, P.P., Itu, C., Scutaru, M.L., Elasto-Dynamics of a
Solid with a General “Rigid” Motion using FEM Model. Part II. Analysis of a
Double Cardan Joint. Vol.58, No.7-8,p.882-892 (2013).
[20] An, Y.C., Wang, X.H., Elastodynamic Analysis of Open Chain Mechanism
with N DOF. SYROM, Bucuresti, Vol.II-1, p.1 (1985).
[21] Deu J.-F., Galucio A.C., Ohayon R., Dynamic responses of flexible-link
mechanisms with passive/active damping treatment. Computers & Structures,
Volume 86, Issues 3–5, pp. 2 (2008).
[22] Fanghella P., Galletti C., Torreg G., An explicit independent-coordinate
formulation for the equations of motion of flexible multibody systems. Mech.
Mach. Theory, 38, p. 417–437 (2003).
[23] Hou W., Zhang X., Dynamic analysis of flexible linkage mechanisms under
uniform temperature change. Journal of Sound and Vibration, Volume 319, Issues
1–2, Pages 570–592 (2009).
[24] Neto M.A., Ambrosio J.A.C, Leal R.P., Composite materials in flexible
multibody systems. Computer Methods in Applied Mechanics and Engineering,
Volume 195, Issues 50–51, p. 6860–6873 (2006).
[25] Shi Y.M., Li Z.F., Hua H.X., Fu Z.F., Liu T.X., The Modelling and
Vibration Control of Beams with Active Constrained Layer Damping. Journal of
Sound and Vibration, Volume 245, Issue 5, p. 785–800 (2001).
[26] Zhang X., Lu J., Shen Y., Simultaneous optimal structure and control design
of flexible linkage mechanism for noise attenuation. Journal of Sound and
Vibration, Volume 299, Issues 4–5, p. 1124–1133, (2007).
[27] Ibrahimbegovic, A., Mamouri, S., Taylor, R.L., Chen, A.J., Finite Element
Method in Dynamics of Flexible Multibody Systems: Modeling of Holonomic
Constraints and Energy Conserving Integration Schemes, Multibody System
Dynamics , Volume 4, Numbers 2–3, 195–223 (2000).
[28] Khulief, Y.A., On the finite element dynamic analysis of flexible
mechanisms. Computer Methods in Applied Mechanics and Engineering, Volume
97, Issue 1, Pages 23–32 (1992).
[29] Zhang X., Erdman A.G., Dynamic responses of flexible linkage mechanisms
with viscoelastic constrained layer damping treatment. Computers & Structures,
Volume 79, Issue 13, p. 1265–1274 (2001).
[30] Vlase, S., Eigenvalues and Eigenmodes of an Inclined Homogeneous Truss
in a Rotational Field. Rom. Journ. Phys., 59, 699-714 (2014).
[31] Sung, C.K., An Experimental Study on the Nonlinear Elastic Dynamic
Response of Linkage Mechanism. Mech. Mach. Theory, 21, p.121–133 (1986).
[32] Alexandre, R.M., Lawrence,K.L., An Experimental Investigation of the
Dynamic Response of an Elastic Mechanism. Journal of REngineering for
Industry, ASME Transaction, February, p.269,(1974).
[33] Biancolini, M.E., Brutti, C., Pennestri, E., Valentini, P.P., Dynamic,
mechanical efficiency and fatigue analysis of the double cardan homokinetic joint,
______________________________________________________________________________________
251
Journal of Vehicle Design, vol.32, n.3/4, pp.231-249 (2003).
[34] Pennestri ,E., De Falco, D., Vita, L., An Investigation of the Inuence of
Pseudoinverse Matrix Calculations on Multibody Dynamics by Means of the
Udwadia-Kalaba Formulation, Journal of Aerospace Engineering, Volume 22,
Issue 4, pp. 365-372 (2009).
[35] Akin, J.E., Application and Implementation of Finite Element Methods.
Academic Press, (1982).
[36] Bathe, K.J., Wilson, E., Numerical Methods in Finite Element Analysis.
Prentice Hall, Inc., (1976).
[37] Cleghorn, W.L., Fenton, E. G., Tabarrok, K.B., Finite Element Analysis of
High-Speed Flexible Mechanism. Mech.Mach.Theory, 16, p. 407 (1981).
[38] Midha, A., ș.a., Finite Element Approach to Mathematical Modeling of High
Speed Elastic Linkage. Mechanism and Machine Theory, Vol. 13, p.603 (1978).
[39] Massonet, CH.,s.a., Calculul structurilor la calculatoarele electronice. Ed.
Tehnică, București (1974).
[40] Erdman, A. G., Sandor, G. N., Oakberg, A., A General Method for Kineto-
Elastodynamic Analysis and Synthesis of Mechanisms. Journal of Engineering for
Industry. ASME Trans., p.1193 (1972).
[41] Gioncu, V., Ivan, M., Basis of the Stability Structures Calculus (in
Romanian). Ed. Facla, Timisoara (1983).
[42] Simeon, B., On Lagrange multipliers in flexible multibody dynamic.
Computer Methods in Applied Mechanics and Engineering, Volume 195, Issues
50-51, Pages 6993-7005 (2006).
[43] Alexandre, R. M., Lawrence, K. L., An Experimental Investigation of the
Dynamic Response of an Elastic Mechanism. Journal of REngineering for
Industry, ASME Transaction, February, p.269,(1974).
[44] Bagci, C., Elastodynamic Response of Mechanical Systems using Matrix
Exponential Mode Uncoupling and Incremental Forcing Techniques with Finite
Element Method. Proceeding of the Sixth Word Congress on Theory of Machines
and Mechanisms, India, p. 472 (1983)
[45] Bagnaru, D., Nanu, G. H., Iordachita, I., Modeles à un nombre fini de
dregrés de liberté pour les mechanisms spatiaux à élements deformables.
SYROM, București, Vol.II-2, p.27 (1985).
[46] Khang N.V., Kronecker product and a new matrix form of Lagrangian
equations with multipliers for constrained multibody systems, Mechanics
Research Communications, Volume 38, Issue 4, pp. 294–299 (2011).
[47] Blejwas, E.T., The Simulation of Elastic Mechanisms using Kinematic
Constraints and Lagrange Multipluiers. Mechanism and Machine Theory, Vol.16,
Nr.4, p.425 (1981).
[48] Blajer W., Kolodziejczyk, K., 2011, Improved DAE formulation for inverse
dynamics simulation of cranes, Multibody Syst Dyn 25: 131–143 (2011).
______________________________________________________________________________________
252
Annex 4C
Numerical prediction of the sound field produced by the
alerting system of EVs and HVs using the BEM
Luís Godinho1, Alba Sánchez-Milan2, Paulo Amado Mendes3, Ramon Peral-Orts4,

Núria Campillo Davó5
1
Luís Godinho, ISISE, Dept. of Civil Engineering, University of Coimbra, Portugal,
lgodinho@dec.uc.pt
2
Alba Sánchez-Milan, Departamento de Ingeniería Mecánica y Energía, Universidad
Miguel-Hernandez, Elche, España, albasanchezmilan@gmail.com
3
Paulo Amado Mendes, ISISE, Dept. of Civil Engineering, University of Coimbra,
Portugal, pamendes@dec.uc.pt
4
Ramon Peral-Orts, Departamento de Ingeniería Mecánica y Energía, Universidad
Miguel-Hernandez, Elche, España, ramon.peral@umh.es
5
Núria Campillo Davó, Departamento de Ingeniería Mecánica y Energía, Universidad
Miguel-Hernandez, Elche, España, ncampillo@umh.es
Abstract. In recent years, both the scientific and the technical community addressed the topic
of noise emission by different types of vehicles. In the particular case of electrical and hybrid
vehicles, the specific problem of near-field sound emission is a concern, since at low speed
these vehicles are considerably silent, and thus much less detectable in a city environment.
Indeed, this topic is of significant importance since it is the means of identifying the presence
of these vehicles by visually impaired persons, and the correct detection within a safe distance
is the only way to minimize possible accidents. The simulation of the sound emitted by these
vehicles can be done using different methods, but in all cases the inclusion of details such as
the directivity of the sound source emitting warning sounds (usually named the Acoustic
Vehicle Alerting Systems - AVAS) is of utmost importance. The Boundary Element Method is
one of the preferred methods to perform acoustic simulation, and is indeed very adequate also
for this type of problems. The present chapter intends to present the BEM as a very good tool
for this purpose. To overcome some limitations of the classic BEM, an “ACA-BEM” model is
here described, and is complemented with adequate Green’s functions to allow the analysis of
some simple configurations of urban environments. Some application examples are then
presented regarding the sound pressure level distribution originated by a possible alerting
system in an open street and in a street with a lateral rigid wall; an addition example is
presented to illustrate the importance of possible obstacles in the propagation path of the
alerting sound, such as the presence of parked vehicles, which indeed act as barriers.
4C.1 Introduction
From an urban environment point of view, one of main characteristics of an Electric

Vehicle (EV) is its absence of mechanical noise, giving as a result very low sound
pressure levels compared to conventional Internal Combustion Engine vehicles (ICE).
But this noise divergence is highly influenced by the speed of the vehicle, on one
______________________________________________________________________________________
253
hand, the difference between the noise level of an EV and an ICE vehicle could be
higher than 6 dB(A) at 10 km/h. However, if the speed is higher than 40 km/h, both
types of vehicles emit similar sound pressure levels, due mainly to the tyre/road noise.
But this characteristic is not thought to be always advantageous, the very low sound
levels due to an approaching electric or hybrid vehicle, masked by the environmental
noise, can be a danger to pedestrians and cyclists. The effect is even worse to
children, elderly and blind people. Some previous studies demonstrated [1], on
different background noise conditions, that the detectability of an EV is significantly
lower than in a conventional ICE vehicles. Along these same lines, some other
publications [2-4] reflect the problem of vehicle detection by visually impaired
persons.
In order to tackle the problem, automotive manufacturers are boarding Acoustic
Vehicle Alerting Systems (AVAS) on EV and HEVs. Regarding that, some
specifications already exist. Japan issued guidelines defining the characteristics of
warning sound in January 2010 and the United States approved legislation on
December 2010. In 2011 the European Commission drafted a guideline for AVAS
and, in 2016, it is expected to be approved the uniform provisions concerning the
approval of Quiet Road Transport Vehicles with regard to their reduced audibility.
This document will stablish that the sound emitted by an EV shall be measured by a
specific methodology in the speed range between 0 km/h and 20 km/h, as well as
specifications on maximum and minimum sound levels emitted by the AVAS. Some
other general characteristics are related with the frequency range. The main
frequencies of the warning sounds should be within 200-1000 Hz and no sound
should be emitted below 100 Hz. The optimal sound system for electric and hybrid
vehicles includes three to four speakers- one in each corner of the car- but in
principle, two loudspeakers would be enough as long as one directs sound forward
and one backwards. However, there is expected any indication related with the sound
directivity of the system, while it is true that it is going to be a relevant characteristic
to optimise the detectability of the vehicle and minimise the annoyance and
environmental impact.
This background information clearly indicates the necessity of performing in-depth
research regarding the use of standardized AVAS in EV and HEV. In this
contribution it is simulated the sound field produced by a mid-size sport utility
vehicle, under different scenarios. In terms of the urban environment, both an open
street and a street with a lateral rigid wall are considered. Regarding the vehicle, two
scenarios are considered in terms of loudspeaker distribution, with one or two front-
placed speakers. Complementarily, the effect of an obstacle (parked vehicle) is also
analysed, introducing further complexity in the sound propagation patterns. The
selected numerical method is the Boundary Element Method (BEM), which is
particularly well suited for acoustic simulations, in which usually an extensive
propagation medium exists around the radiating sources. However, standard BEM
formulations can pose significant limitations regarding problem size, and special
strategies must be used whenever larger problems need to be analysed. In the present
work, an Adaptive-Cross-Approximation algorithm is used, as described [5], and
previously applied by the authors [6].
______________________________________________________________________________________
254
4C.2 Governing equations for acoustic wave propagation
The propagation of sound within a three-dimensional space can be mathematically

represented in the frequency domain by the Helmholtz partial differential equation as:
NS
∇ 2 p + k 2 p = −∑ Qk δ (ξ kf , ξ ) Eq. 1 - C
k =1
where ∇2 = ∂ 2 ∂x 2 + ∂ 2 ∂y 2 + ∂ 2 ∂z 2 , p is the acoustic pressure, k = ω c ,

ω = 2π f , f is the frequency, c is the sound propagation velocity within the
acoustic medium, NS is the number of sources in the domain, Qk is the magnitude of
the existing sources ξ kf located at (xξkf

, yξ f , zξ f
k k
) , ξ is a domain point located at
( x , y , z ) and δ (ξ
ξ ξ ξ k
f
, ξ ) is the Dirac delta generalized function.
In the above defined Helmholtz equation, the Sommerfeld radiation condition
( ) ( )
lim ⎡⎣ ∂ p X ∂n − ikp X ⎤⎦ = 0 is automatically satisfied at infinity, where X is the
x→0
field point located at ( x, y, z ) , n is the unit outward normal vector and i = −1 .

Considering that a point source is placed within this propagation domain, at
X 0 = ( x0 , y0 , z0 ) , it is possible to establish the fundamental solution for the sound
2 2 2
pressure at a point X , at a distance r = ( x − x0 ) + ( y − y0 ) + ( z − z0 ) , which can
be written as:
e − ikr
G (ξ , X ) = Eq. 2-C
4π r
This simple fundamental solution can be further developed in order to account for
some particular environmental configurations of the propagation medium.
Considering an acoustic analysis, the presence of perfectly reflecting plane surfaces
can be taken into account by using the well-known image-source method. In this
technique, the effect of a point source in the presence of a given plane surface can be
simulated by considering an additional virtual source, positioned in a symmetrical
position with respect to the reflecting plane. Thus, following, for example, [7] if such
plane is defined by z = 0 , the corresponding Green’s function can be written as
− ikr
e − ikr e 1
Ghalf (ξ , X ) = + Eq. 3-C
4π r 4π r1
______________________________________________________________________________________
255
2 2 2
where r1 = ( x − x0 ) + ( y − y0 ) + ( z + z0 ) . The above expression is only valid if
the plane is perfectly rigid, with a reflection coefficient of 1, although these
expressions can be easily adjusted to include a different reflection coefficient. More
complex configurations may also be simulated, as described in [7].
z
z
y y
x x
a) b)
Fig. 1-C. Some simple configurations simulated using image-sources: a) half-space; b) quarter-
space.
4C.3 The Boundary Element Method - BEM
The Boundary Element Method is a boundary technique, in which the relevant

quantities and calculations are performed over the problem’s boundary. To formulate
the method, consider Green’s Second Identity applied to Eq. 1-C, originating the
following Classic Boundary Integral Equation (CBIE):
∂G (ξ , X )
C (ξ ) p (ξ ) = −iρω ∫ G (ξ , X ) vn ( X )dΓ − ∫ p ( X ) dΓ +
∂n
Γ Γ
Eq. 4-C
NS
+ ∑ Qk G (ξ , ξ )k
f
k =1
(
where Γ is the boundary surface, ρ is the medium density, G ξ kf , ξ is the incident )
field regarding the acoustic pressure generated by the real source placed at position
ξ kf ; p ( X ) and vn ( X ) represent the acoustic pressure and the normal component of
the particle velocity at X , respectively. The coefficient C (ξ ) depends on the
boundary geometry at the source point, and equals 0.5 for a smooth boundary.
For the case of exterior problems, it is known that equation 4-C provides unstable
solutions at certain frequencies. Strategies have thus been developed in order to avoid
this significant drawback, namely by means of its combination with the so-called
Hypersingular Boundary Integral Equation (HBIE). To obtain this second equation,
the first derivative with respect to the normal direction at the source point, nL , must
be considered, and the following equation arises:
______________________________________________________________________________________
256
( ) ( )
−iρω C! ξ vn ξ = −iρω ∫
(
∂G ξ , X ) v ( X )dΓ −
Γ
∂nL n
Eq. 5-C
(
∂2 G ξ , X ) p ( X ) dΓ + NS
(
∂G ξ kf , ξ )
∫ ∂nL ∂n
∑Q k
∂nL
Γ k=1
As before, the coefficient C%(ξ ) depends on the boundary geometry at the source
point, and equals 0.5 for a smooth boundary.
It has been proposed by Burton and Miller [8] that combining the two equations 4-C
and 5-C in the form
CBIE + i/k HBIE = 0 Eq. 6-C
originates a new boundary integral equation which provides unique solutions at all
frequencies. Different boundary conditions may be prescribed at each boundary
element, such as Dirichlet ( p ( X ) = p ( X ) ), Neumann ( v ( X ) = v ( X ) ) or Robin (
p ( X ) v ( X ) = Z ( X ) , being Z ( X ) the acoustic impedance) boundary conditions.
After prescribing such conditions, discretizing the boundary into NEB segments, and
establishing adequate interpolation functions within each segment (constant
interpolation in the present work), a system with NEB equations on NEB unknowns
can be assembled. Its solution makes it possible to obtain the acoustic pressure and
the normal velocity at each boundary element, and consequently the pressure at any
point of the domain can be computed by applying the boundary integral Eq. 4 - C.
4C.4 Adaptive Cross Approximation formulation
One of the most significant drawbacks of the classic BEM is the difficulty to address
problems with a very large number of elements, and thus to tackle realistic
engineering problems, without making use of additional mathematical strategies.
Indeed, the BEM generates fully populated matrices, and the related memory
requirements grow quadratically with NEB . An estimation of the required computer
memory in acoustic problems (with complex numbers and double precision) can be
obtained simply as memory( MB) = 16 NEB 2 / 10242 , and it is thus simple to
understand that with as little as 10000 elements, 1.5 GB of RAM are required just for
the system matrix. Fig. 2-C illustrates the growth of the memory requirements of the
BEM with the number of nodes.
______________________________________________________________________________________
257
Fig. 2-C. Memory requirements of the classic BEM as a function of the number of nodes.
To overcome this difficulty, different strategies have been developed by the scientific
community [5]. Among these, perhaps the most widely known is the Fast-Multipole
method, which allows tackling problems with more than 1 million nodes. However,
mathematically it involves some substantial changes to the way the BEM is
implemented, and thus the reuse of existing codes is more difficult.
An interesting alternative for mid-sized problems is the use of the Adaptive Cross
Approximation technique [5,6]. The Adaptive Cross Approximation (ACA) is based
in hierarchical matrices (H matrices), an approach that allows for matrix operations
with almost linear complexity. These matrices may be understood as an algebraic
structure reflecting a geometrically motivated partitioning into sub-blocks, that in turn
are classified to be either admissible or not. This block structure points out the fact
that the H-matrix arithmetic is easily parallelizable. After having concluded the setup
of each block, admissible blocks have to be approximated. The main advantage of
ACA is that, for such approximation, it only requires the evaluation of some original
matrix entries and the approximation is still almost optimal. Having a matrix A ∈ C t × s
, the approximation is obtained as follow:
A ≈ Sk = UV T Eq. 7-C
where U ∈ C t ×k and V ∈ C s×k . In this approximation k has a smaller rank when
compared with t and s . This low-rank approximation can be obtained only for well-
separated computational domains x ≠ y .
To exploit the full advantages of such approach, it becomes necessary to solve the
corresponding equation system by means of an iterative solver. In the examples
presented here, the Generalized Minimal Residual Method (GMRES) iterative solver
has been used, together with a block preconditioner (to accelerate convergence) based
on the non-admissible blocks of the matrix.
______________________________________________________________________________________
258
4C.5 Sound field calculation around emitting vehicles
To illustrate the applicability of the described ACA-BEM in the analysis of the sound
field around an emitting EV or HV, in this section a number of simple examples are
presented. A very simple geometry of the vehicle is chosen just as an illustrative case,
as depicted in Fig. 3-Ca.
In a first set of examples, let us consider a simple vehicle emitting a warning sound
from a single frontal loudspeaker in an open street. The assemblage of the BEM
system matrix is performed as described in the previous section, considering an ACA
algorithm, in which admissible blocks are greatly compressed. Fig. 3-Cb illustrates
the hierarchical matrix constructed for this purpose, where white patches correspond
to admissible blocks (composing the vast majority of the matrix). Comparing BEM
and ACA-BEM memory requirements, it can be stated that in ACA-BEM the required
memory is reduced to 35%.
a) b)
Fig. 3-C. Discretized vehicle using 9056 triangular boundary elements (a) and corresponding
ACA-BEM hierarchical matrix representation (b).
For this geometry, sound pressure levels have been calculated considering the single
loudspeaker to be placed in the front part, centred with the vehicle. In Fig. 4-C, the
sound pressure levels in a plane parallel to the ground is illustrated for 250Hz, 500Hz,
750Hz and 1000Hz. It is clear that the increasing excitation frequency originates more
intricate SPL maps, with the interaction between direct sound waves and the various
reflections being responsible for this complexity. In addition, regions with very low
SPL are visible at varying distances, which are generated by the interference between
the direct waves and the first reflection on the rigid ground.
______________________________________________________________________________________
259
a) b)
c) d)
Fig. 4-C. Results for 250Hz (a), 500Hz (b), 750Hz (c) and 1000 Hz (d). Colour scale from
15 dB (dark blue) to 75 dB (dark red).
To have a better and clearer view of the observed behaviour, polar plots are depicted
in Fig. 5-C, considering a distance from the front of the vehicle of 10 m, and for
receivers placed 1.6 m above the ground. As stated in the introduction, the distance of
10 m can be considered as the minimum safe distance for vehicle detection by
pedestrians, and thus it is worth a more detailed analysis.
The polar plots in Fig. 5-C evidence a distinct directivity pattern, varying with the
excitation frequency. For lower frequencies, in spite of the presence of the rigid
vehicle shadowing the source, relatively high SPL is reached for all angles, although
with somewhat higher values in the front part (a difference of 5 to 10 dB). For higher
frequencies, and as in Fig. 4-C, a more complex behaviour is observed. It is
interesting to note, for the frequency of 750 Hz, a very significant distinction between
the front and rear parts of the vehicle, with energy being more concentrated in the
front part, as expected.
It is important to note that, in the previous simulations, the emitting source was
assumed to be a point source placed near the front panel of the vehicle, approximately
0.1 m away from it. To better understand the influence of this simplification, a second
set of simulations was performed embedding this source within the frontal panel,
introducing a small semi-spherical cavity where the source is located. A detailed view
of the corresponding BEM mesh is depicted in Fig. 6-Ca, and the computed polar
plots, indicating the directivity of the system, are illustrated in Fig. 6-Cb, for a
frequency of 750Hz. Comparing with the plots of Fig. 5-C, some differences are
visible, particularly in what concerns the rear part of the vehicle, where less energy
seems to be arriving; by contrast, in the front part, an increase in the SPL is visible,
and was to be expected due to the more intense shadowing/baffle effect originated by
the cavity where the source is now placed. Remarkably, the main features of the
______________________________________________________________________________________
260
directivity plot in the frontal part are almost exactly maintained, indicating that, for
the purpose of evaluating the effect of the source in the front region, making use of
the simpler approach of Fig. 5-C can be sufficient.
a) b)
c) d)
Fig. 5-C. Polar plots for 250Hz (a), 500Hz (b), 750Hz (c) and 1000 Hz (d).
a) b)
Fig. 6-C. Effect of considering the source embedded in the front panel: (a) detailed view of the
BEM mesh and (b) polar plot for a frequency of 750 Hz.
In the previous examples, a single emitting speaker was considered in the front of the
vehicle. However, it has been proposed by different researchers that better
detectability may be obtained if more sources are placed emitting the warning sound.
______________________________________________________________________________________
261
In the plots of Fig. 7-C, the directivity pattern is compared in polar plots for situations
with one and two acoustic sources emitting noise in the frontal part of the vehicle. As
expected, the polar plot for the latter case indicates a more dense concentration of
energy in front and to both sides of the vehicle; indeed, it is interesting to note that the
directivity plot exhibits two broad lobes, one to each side of the vehicle, where
significant and homogenous SPLs are attained, indicating that a significant area is
illuminated by the emitting system.
a) b)
Fig. 7-C. Effect of one (a) or two (b) emitting sources in the front part of the EV, for 750Hz.
In what follows, two distinct situations, usually occurring in an urban environment,

are now considered, namely the presence of a parked car (obstacle) and the presence
of a rigid wall parallel to the street.
Let us start by analysing the effect of a right-sided parked vehicle, acting as a barrier
to the propagation of the warning sounds emitted by the AVAS. Fig. 8-C illustrates
both the SPL and the directivity polar plot for such situation. The SPL colour map of
Fig. 8-Ca clearly indicates the existence of a shadow effect behind the parked vehicle,
which, in practice, acts as a barrier to the propagation of sound waves. Observing the
polar plot in Fig. 8-Cb, this effect is clearly seen, with a significant decrease of the
energy reaching the region where pedestrians circulate. Although this decrease is
largely dependent on the relative position of the two vehicles, the presented result
gives a clear indication regarding a potential acoustic shadowing effect that may
decrease detectability of the EVs.
Results in Fig. 9-C refer to the last analysed case, in which a rigid lateral wall is
considered parallel to the street. Results for 1000Hz are illustrated. The results,
presented in terms of SPL maps, indicate the presence of a reflected field coming
from the rigid vertical wall, originating a more complex sound field. This reflection
may enrich the sound field and may slightly increase the detectability of the vehicles
by pedestrians. However, it must be stressed that this second wave is traveling from a
direction opposite to the car (from the wall side and not from the street side), and can
thus induce some confusion as to where the vehicle is situated (particularly in the case
of visually impaired people). This behaviour merits a more in-depth study, and falls
beyond the scope of the present chapter.
______________________________________________________________________________________
262
a) b)
Fig. 8-C. SPL distribution (a) and polar plot (b) calculated in the presence of a parked car, for a
frequency of 750Hz.
Fig. 9-C. SPL distribution calculated for a frequency of 1000Hz: a) open street; b) street with
a rigid lateral wall.
4C.6 Final remarks
This chapter addressed the potential of modern numerical methods in the prediction of
the sound field generated by circulating EVs and HVs in an urban environment. The
specific case of the Boundary Element Method (BEM) has been focused, presenting
its mathematical formulation and addressing a possible simple strategy adequate to
allow tackling mid-sized problems (up to several hundred thousands of nodes). This,
technique, the “ACA-BEM”, complemented with adequate Green’s functions, was
here used to illustrate some numerical examples regarding the study of the effect of
AVAS used in electrical and hybrid vehicles in an urban environment. Open and
straight-shaped configurations and the presence of other vehicles (acting as barriers)
were here illustrated. Although the objective of the chapter was just to illustrate the
applicability of the numerical tool, it was also possible to draw interesting conclusions
regarding the global acoustic behaviour of the system. It was observed that the effects
of parked vehicles and the specific urban configuration can greatly influence the
______________________________________________________________________________________
263
capacity of pedestrians to detect circulating vehicles. To finish, it should be noted that
further research in this topic is required, and objective quantification of the
detectability of these vehicles deserves further attention.
4C.7 References
[1] Garay-Vega, L., Hastings, A., Pollard, J. K., Zuschlag, M., & Stearns, M. D.
Quieter cars and the safety of blind pedestrians: Phase I (No. HS-811 304) (2010).
[2] Verheijen, E. Jabben, J. Effect of electric cars on traffic noise and safety. Report
680300009/2010. National Institute for public Health and the Environment
Netherlander (2010).
[3] Goodes, P., Bai Y. B., Meyer, E., Investigation into the Detection of a Quiet
Vehicle by the Blind Community and the Application of an External Noise
Emitting System. SAE International. 2009-01-21895 (2009).
[4] Mendonça, C., Freitas, E., Ferreira, J.P., Raimundo, I.D., Santos, J.A. Noise
abatement and traffic safety: The trade-off of quieter engines and pavements on
vehicle detection. Accident Analysis and Prevention 51(2013) 11-17.
[5] Liu Y. J., Mukherjee S., Nishimura N., Schanz, M., Ye, W., Sutradhar, A., Pan,
E., Dumont, N. A., Frangi, A., Saez, A. Recent Advances and Emerging
Applications of the Boundary Element Method. Appl. Mech. Rev. 64-3 (2012)
030802.
[6] Godinho, L., Amado-Mendes, P., Santos, P. G., Cordeiro, M. An efficient and
accurate numerical model for the 3D analysis of sonic crystals. 22nd International
Congress on Sound and Vibration, Florence (2015).
[7] Godinho, L., Costa, E., Santiago, J., Pereira, A., Amado-Mendes, P. 3D analysis
of the sound reduction provided by protective surfaces around a noise source.
International Journal of Acoustics and Vibrations 19-2(2014) 78-88.
______________________________________________________________________________________
264
Annex 4D
Equivalent material modelling of sandwich beam
assemblies: propagating and evanescent wave
considerations
Sophie de Rijk1, Eugéne Nijman 1
1
Virtual Vehicle Research Center, Inffeldgasse 21/A, 8010 Graz, Austria,
sophie.derijk@v2c2.at, eugene.nijman@v2c2.at
Abstract. An approach to modelling sandwich structures in bending with a homogenised

equivalent material is presented. The dispersion of the layered system is brought to a
Timoshenko behaviour, and an equivalent damping coefficient is derived. The equivalent
model can then be implemented in finite element (FE) models by means of commonly used
shell elements with transverse shear stiffness. A critical review of the limitations of such an
equivalent material is made, specifically in relation to the evanescent parts of the dispersion,
which are of importance in assembled structures. A benchmark case with specific boundary
conditions is presented to demonstrate the strengths and weaknesses of the approach.
4D.1 Introduction
In an effort to build more fuel-efficient vehicles, high stiffness to mass ratio solutions
for vehicle structures are being researched and used in the automobile, train and
aerospace industries alike. This is in particular becoming critical for vehicles
equipped with hybrid or electric powertrain in order for the additional weight of the
battery cells and the auxiliary units to be leveraged. However, the introduction of light
yet very stiff composite materials in automotive structures has an important
implication on their vibro-acoustic behaviour. The high stiffness-to-weight ratio
namely heavily impacts the structural dynamics and acoustical properties of these
structures leading to novel, emerging noise, vibration and harshness (NVH)
challenges. In particular, the air-borne sound insulation properties become very poor
compared to conventional materials. This is due to the (i) high stiffness-to-weight
ratio implying low coincidence frequency and (ii) the inherently low mass, which is
however the governing parameter when it comes to air-borne insulation in the low
frequency range. The resulting trade-off between the reduced weight and worsened
NVH must be therefore accounted for. Furthermore, hybrid and electric vehicles
exhibit an acoustic footprint different to that of vehicles equipped with conventional
internal combustion engine, which implies specific requirements to be considered
during the product development phase.
In this context, sandwich constructions have very promising characteristics as they
present high bending stiffness for very low mass. Such structures however show a
variety of types of behaviours over the range of frequencies in which they are excited
during operation. Calculating their response by representing all their layers in FE
______________________________________________________________________________________
265
requires large computing power and time. For design and modelling purposes,
specifically in FE, representing the multi-layered system with an equivalent single-
layered material would make computing costs significantly lower and allow for much
more flexibility during the development phase of vehicles. Since the aim is to set-up a
methodology easily employable using commercial FE packages, making an
equivalent material using Timoshenko beam theory would be a convenient choice as
it can be implemented easily using shell elements. In this paper we first present a brief
overview of existing methodologies for modelling sandwich structures with
equivalent beams, including damping and shear modelling in laminates.
Subsequently, an example case is presented and the issues and shortcomings of the
approach are highlighted.
4D.2 State of the art in sandwich beams and plates dynamics
Many publications can be found on vibrations of sandwich beams and plates. Over the
past few decades, researchers have strived to describe the vibrations in bending of
sandwich structures using theories with complexities ranging from classic plate theory
(Euler beam), first order shear (Mindlin for plates, Timoshenko for beams) to higher
order theories. Several reviews concerning previous work on vibrations of laminates
and sandwich structures already exist. We redirect the reader to, for instance, the
review of Mallikarjuna and Kant [1] or that of Denli and Sun [2]. This section limits
itself to giving some examples of typical approaches that can be found in the
literature. For general orthotropic laminates, Yang, Norris and Stavsky [3] developed
general equations of motion, assuming plane stress in each layer and first order shear
theory for the global displacements of the plate. They then also simplify to classic
plate theory by ignoring transverse shear and rotary inertia terms. The stress-strain
relations of the laminate are obtained by integration of the properties of each layer
over the cross section. Yu and Cleghorn [4] calculate the eigenfrequencies of simply
supported symmetric sandwich plates by constructing the stiffness and mass matrices
of the system following in turn the classic plate theory, the Mindlin plate theory and
Reddy’s third order plate theory. The stress-strain relations are calculated analogously
to [3]. Lok and Cheng [5] obtain the various stiffness terms for a Mindlin plate from
static stiffness calculations of a unit element of the sandwich structure and then
calculate their eigenfrequencies assuming simply supported boundary conditions.
Another approach, presented for instance by Mead [6], Zhou and Li [7], and Banerjee
et al. [8], is to consider separately the motion of the face sheets and core, using
different assumptions for each layer and then coupling the layers in one plate.
Publications by Gaudenzi and Carbonaro [9] and Kant and Swaminathan [10] are
examples of higher order theory applications. Backström [11] bases his work on the
description of the behaviour of a sandwich beam described by Nilsson and Nilsson
[12] to define frequency-dependent stiffness terms for either an equivalent Euler or
Timoshenko beam.
______________________________________________________________________________________
266
Damping
When modelling a sandwich structure as a single material, the various loss factors of
each layer need to be merged into a single equivalent loss factor. An often adopted
assumption in the literature is that only shearing in the core is dissipative since it
typically has higher damping but relatively low elastic modulus, see [6]. Sadasiva,
Rao and Nakra [13] consider the core to be dissipative both in in-plane elongation and
transverse shear motion and calculate a system loss factor as the ratio between the
imaginary and real parts of the complex stiffness for simply supported beams. Maheri,
Adams and Hugon [14] treat the case of a simply supported plate, considering
damping for the in-plane elongation, in-plane shear and transverse shear motion. The
in-plane stresses in the core and transverse shear in the face sheets are however
neglected. Modal damping capacities are then obtained by solving the eigenvalue
problem. Cremer, Heckel and Petersson [15] present a more generalised methodology
to compute system loss factors of layered systems by defining it as the ratio of
dissipated energy over potential strain energy per cycle.
Transverse shear stress distribution and correction factors
In the formulations given by the first order shear theory as well as the one given by
Nilsson and Nilsson [12], the shear term Ds implies uniform shear stress distribution
over the cross section. This is however inaccurate as the shear stiffness varies over the
layers and a no-shear condition is required at the top and bottom surfaces of the beam.
To account for this, a shear correction factor κs must be applied to Ds. In the case of
beams with rectangular cross section and homogeneous material, the default reference
shear correction factor applied in most FE packages (for example ABAQUS or
NASTRAN) for shell elements is κs = 0,833 given by Reissner [16], assuming a
parabolic stress distribution. Several authors have also addressed the issue of defining
such a correction factor for laminates, see for example [17], [18], [19] and [20]. The
shear correction factor is calculated such that the shear strain energy from the uniform
stress/strain distribution and the actual stress/strain distributions are the same. The
difference in the approach between the various authors lies in the assumptions on the
actual stress distribution.
4D.3 From Nilsson sandwich beam to Timoshenko equivalent

beam
For the work presented in this paper, the reference theory for the motion of a
sandwich beam is taken from Nilsson and Nilsson [12]. This theory considers the total
bending of the layup, the transverse shearing of the core and the individual bending of
the face sheets about their own mid plane as well as global transverse and rotary
inertias. The governing equations are then derived using the Hamiltonian approach:
______________________________________________________________________________________
267
!! ! !" !! ! !! ! !! !
−𝐷! − + 2𝐷! − +𝜇 =0 Eq. 1-D
!! ! !" !! ! !! ! !! !
! ! ! !
! ! ! ! ! ! !" ! !
−𝐷!"! + 2𝐷! − − 𝐷! − 𝜓 + 𝐼! =0 Eq. 2-D
!! ! !! ! !! ! !" !! !
where W the transverse displacement and ψ the rotation angle of the cross sections.
From Eq. 1-D and 2-d and considering a general solution of the form 𝑊 = 𝑒 !(!"!!") ,
a 6th order dispersion equation is obtained:
!! !! !! !! !!! !"!
2𝐷! 𝜅 ! + 𝐷! − 2 𝐼 𝜔! 𝜅! − 𝜇 + 2 𝜇+ 𝜅 ! 𝜔! − 𝜔! + 𝜔 ! = 0 Eq. 3-D
!!"! ! !!"! !!"! !!"! !!"!
where Df is the individual bending stiffness of the face sheets, Ds the shear stiffness,
Dtot the total bending stiffness, Iρ the rotary inertia and µ the mass per unit length. The
shear stiffness is considered to be simply that of the core, neglecting the contribution
of the face sheets. We also point out here that in the original paper, no shear
correction factor is considered. From Eq. 3-E we obtain three distinct wave numbers.
In the undamped case, the first wavenumber κ1, is strictly real and represents the
propagating part of the solution going from pure bending at low frequency to shearing
in the mid-frequency range, then over to individual bending of the face sheets about
their own mid plane in the high frequency range. The second wavenumber κ2, takes
purely imaginary values in the lower frequency range and real values past a given
limit frequency. Below the limit frequency, the second wavenumber represents a
slowly decaying evanescent solution. Above the limit frequency, the solution is
propagating. The third wavenumber κ3, is purely imaginary and is a very local
exponential solution with a short decay. The various solutions of the dispersion
relation (Eq. 3-D) are shown in Fig. 1-D.
Fig. 16-D. Solutions to the dispersion relation following Nilsson and Nilsson [12] for a 15 mm
thick aluminium sandwich beam with 1 mm face sheets
As mentioned in the introduction, we will model sandwich beams with an equivalent

Timoshenko beam for which the governing equations are:
!! ! !" !"
−𝐷!" − + 𝜇! =0 Eq. 4-D
!" ! !" !! !
______________________________________________________________________________________
268
!! ! !" !! !
−𝐷!" − 𝐷!" − 𝜓 + 𝐼!" =0 Eq. 5-D
!" ! !" !" !
From which we obtain the following 4th order dispersion equation:
!!" !!" !!" !!

𝐷!" 𝜅 ! − 𝜇! + 𝐼!" 𝜅 ! 𝜔 ! − 𝜇! 𝜔! + 𝜔! = 0 Eq. 6-D
!!" !!" !!"
where µt is the mass per unit length, Iρt the rotary inertia, Dst is the transverse shear
stiffness and Dbt the bending stiffness. We can then define an equivalent material by
choosing an elastic modulus, shear modulus, density and total thickness such that Dbt
= Dtot, Dst = Ds, Iρt = Iρ and µt = µ. We draw attention to the fact that, while all
previous authors agree on matching the bending and shear stiffness and density, most
authors define the thickness of the equivalent material (Ht) to be either the total
thickness of the sandwich layup, or that of the core alone. We choose to match the
rotary inertia instead, requiring the thickness of the equivalent material to be different
from the total thickness of the layup:
𝜇! = 𝜇 !"!! !
𝐼!" = 𝐼! 𝐻! = , 𝜌! = Eq. 7-D
! !!
This choice is driven by the fact that, while the rotary inertia is typically too small to
influence the propagating part of the solution κ1 of the sandwich and Timoshenko
beams, the κ2 part of the solution is sensitive to the accuracy of Iρ. As shown in Fig. 2-
D, if Iρ is not exactly matched (but the total thickness Htot is), the transition frequency
between evanescent and propagating solutions of κ2 changes as well as the
propagating κ2, compared to the sandwich dispersion.
Fig. 17-D. Dispersion solutions for a 15 mm aluminium honeycomb sandwich beam with 1 mm
face sheets compared to equivalent Timoshenko solutions either matching Htot and µ or Iρ and µ
______________________________________________________________________________________
269
Equivalent loss factor
When using shell elements in FE, one typically sets the damping with a single loss
factor such that 𝐸! = 𝐸!∗ (1 + 𝑖𝜂!"! ) and 𝐺! = 𝐺!∗ (1 + 𝑖𝜂!"! ), where 𝐸!∗ and 𝐺!∗ are the
undamped moduli. We therefore need to define a single equivalent loss factor from
the various loss factors of the face sheets and the core. We follow the approach given
in [15] where the total loss factor 𝜂!"! is calculated as the ratio between dissipated
energy and potential strain energy:
!!"##
𝜂!"! = Eq. 8-D
!!!!"#
We then consider the dissipative parts of the motion to be the elongation in the face
sheets due to bending and the transverse shearing in the core:
!!! !
𝐸!"## = 𝐸!"##$ + 𝐸!"##$ = 2𝜋𝜂! 𝐸! ℎ! + 𝜋𝜂! 𝐺!∗ 𝐻! 𝛾 !
Eq. 9-D
!"
where ℎ! the thickness of the face sheets, 𝐻! the thickness of the core and 𝜖! the
longitudinal displacement of the face sheets due to bending, which can be expressed
!! !!!
in function of the rotation of the cross section 𝜓: 𝜖! = 𝜓. 𝛾 is the shear angle
!
!"
given by: − 𝜓. The strain potential energy is considered, as given in [12], to be the
!"
sum of the total bending energy, the individual face sheet bending energy of the face
sheets and the transverse shear energy:
! !" ! !" ! ! !
𝐸!"# = 𝐷!"! + 𝐷! + 𝐷! 𝛾 Eq. 10-D
! !" !" !
Using the governing equations of the Nilsson sandwich beam, we can express 𝛾 in
function of 𝜓. Considering the general solution of the form 𝜓 = 𝑒 ! !"!!" , deriving
by 𝑥 is equivalent to multiplying by the wavenumber 𝜅 (we consider the propagating
𝜅 (we consider the propagating 𝜅! ):
!! !!"! !! !" !!"! !!

𝜅; 𝛾 𝜅! + 𝜅! − 1 ; 𝜅! + 𝜅! − 𝜅 Eq. 11-D
!" !!! ! !" !!! !
The total equivalent loss factor is then:
! ! !!
! ! ∗ ! (! !!! )! !! !!! !!∗ !! ( !"!
! ! ! ! !
!! ! ! !! !!)!
!! !
𝜂!"! = ! !! ! !! Eq. 12-D
!!"! !! !!!! !"! !! ! ! !! !! !!! ( !"! !! ! ! !! !!)!
!! ! !! !
______________________________________________________________________________________
270
Slider-clamped beam numerical example
In the following, we calculate the response of a sandwich beam and its Timoshenko
counterpart using the analytical formulations given in Section 3. All computations are
implemented in MATLAB. For this numerical example, we choose to model a beam
which is loaded by a unit force at 𝑥 = 0 where a slider condition is assumed, so
transverse motion 𝑊 is allowed and is clamped at 𝑥 = 𝐿 = 600 𝑚𝑚. The boundary
conditions set for both beam theories are listed in Table 1-D.
Table 1-D. Considered boundary conditions
The sandwich beam modelled is composed of 1 mm thick aluminium face sheets

!"
(𝐸! = 70 𝐺𝑃𝑎, 𝐺! = 0, 𝜌! = 2730 ! , 𝜂! = 0,01) Following Nilsson [12], the
!
bending stiffness is set as that of the whole laminate, the shear stiffness to be given
solely by the core with no shear correction factor. The rotary inertia and density per
unit length are also kept identical. The damping of the equivalent material is defined
in two different ways. The first, which is only possible if we are computing the
response analytically, consists in setting separately the loss factor of the face sheets
(on the bending terms) and of the core (in the shear terms), identically to the sandwich
formulation. It therefore represents a best case scenario. The second is defined
following Eq. 12-D as a unique equivalent loss factor. Fig. 3-D shows the input
mobility and dispersion solutions of the sandwich with its Timoshenko counterpart.
Fig. 4-D then shows the force and moment transfer mobilities obtained in this case.
While Fig. 3-D reflects the accuracy of the global response of the equivalent material,
Fig. 4-D is meant to show how the boundary conditions are respected by the
equivalent material
______________________________________________________________________________________
271
Fig. 18-D. Dispersion and input mobility for a 15 mm sandwich beam and its equivalent
Timoshenko beam, with damping defined by separate loss factors or by equivalent loss factor
Fig. 19-D. Force (top) and Moment (bottom) transfer impedance between x=0 and x=L for a 15
mm sandwich beam and its equivalent Timoshenko beam, with damping defined by separate
loss factors or by equivalent loss factor
Firstly we notice from Fig. 3-D and 4-D alike that the equivalent material with the
damping defined with separate loss factors follows very accurately the behaviour of
the sandwich beam up to about 𝑓 ≅ 3 𝑘𝐻𝑧 (with the exception of the first peak
value). Further on, a slight shift in the frequencies can be noticed (about 35 Hz shift at
3 kHz). The peak values however remain very close to the sandwich values (within
0.2 dB). At about 𝑓 ≅ 6 𝑘𝐻𝑧 the response given by the equivalent material starts to
diverge clearly from the reference sandwich. Looking at the response of the
equivalent material with the unique global loss factor, we notice that it clearly
______________________________________________________________________________________
272
underestimates the damping, but its accuracy improves as the frequency increases.
The equivalent global loss factor is defined using only the propagating part 𝜅! of the
dispersion solution. At low frequencies however, the wavenumber of the evanescent
κ! is of the same order of magnitude as that of the propagating solution 𝜅! . This
implies that the evanescent field affects the total displacement (and therefore
dissipated energy) with similar wavelengths as the propagating solution. As the
frequency increases, the wavenumber of the propagating solution increases, which
means it increasingly dominates the total integrated dissipated energy, while the
evanescent solution serves more as a local scaling. This is visible in both Fig. 3-D and
4-D, whereas frequency increases, the Timoshenko beam with the total equivalent
loss factor starts to converge to the solution of its counterpart with the damping
defined by parts. This observation highlights the importance of the evanescent field at
lower frequencies and should therefore not be ignored. The influence of inaccurate
damping is all the more striking in the force and moment transfer impedances.
Another issue concerns the representation of shear stiffness. In the (common) case of
thin face sheets, the assumption that the shear stiffness is only given by the core is
reasonable. However, setting a shear correction factor of 1 is, in such cases, also
inaccurate. The shear stress distribution in the laminate doesn’t merely depend on the
shear stiffness terms but also on the in-plane stiffnesses of the layers [17][18][19][20].
As the face sheets have high in-plane stiffness, their presence modifies the shear
stress distribution. For the laminate considered here, the shear stiffness of the face
sheets if of ℎ! 𝐺! = 2,63 10! , that of the core is of 𝐻! 𝐺! = 2,72 10! , if one calculates
the total shear stiffness following the approach of Madabhusi-Raman and Davalos
[20] neglecting cross-plane (plate) effects, the total shear stiffness of the laminate
becomes 𝐷! = 3,16 10! . In this case, the difference in stiffness is of an acceptable
16%, but it clearly indicates that the face sheets do have an influence on the total
shear stiffness. In order to be as generic as possible and cover a wide range of
laminates from thinner to thicker face sheets, our methodology should include the
face sheets in the computation of the shear stiffness. Simply adding up the stiffnesses
is however clearly not an option, the actual stress distributions needs to be considered
via shear correction factors. If a fully generic approach is considered in which one
would include the face sheet shear stiffness, an added complication then comes when
considering damping. As shearing is then considered over both the face sheets and
core, the energy dissipated in each part needs to be quantified separately. This
requires defining multiple shear correction factors for the laminate, one for the face
sheets, one for the core:
!"! !
𝐸!"## = 𝐸!"##$ + 𝐸!"##$ = 2𝜋𝜂! 𝐸!∗ ℎ! + 𝜋(2𝜅!" 𝜂! 𝐺!∗ ℎ! + 𝜅!" 𝜂! 𝐺!∗ 𝐻! ) 𝛾 !
Eq. 13-D
!"
where 𝜅!" and 𝜅!" are the shear correction factors in the face sheets and the core.
These two shear correction factors can be obtained following the methodology given
by Vlachoutsis [19] such that:
! !
𝜅! !!! 𝐺! ℎ! = !!! 𝜅!" 𝐺! ℎ! Eq. 14-D
______________________________________________________________________________________
273
were 𝐺! , ℎ! 𝑎𝑛𝑑 𝜅!" are the shear modulus, the thickness and the shear correction
factor of each layer respectively.
4D.4 Conclusions
In this paper, the approach to building an equivalent Timoshenko beam for

representing a sandwich laminate based on current knowledge is presented. The final
aim is to accelerate FE calculations of sandwich structures using shell elements with
equivalent material properties. Then, on the basis of a practical example, some of the
challenges that need to be tackled with ongoing research to make a reliable and
generically applicable equivalent model are highlighted. The main issue arising from
this example is the proper definition of an equivalent loss factor to be applied to the
equivalent material. It has been shown here that the latter depends not only on the
propagating part of the solution, which is commonly considered by modelling simply
supported structures (purely sinusoidal solution), but also the evanescent field,
especially at lower frequencies. The necessity to compute the shear stiffness
considering both the face sheets and the core as well as separate shear correction
factors is also pointed out.
Acknowledgements
The research work of Sophie de Rijk has been funded by the European Commission
within the Marie Curie program through the EID eLiQuiD project (GA 316422). The
authors acknowledge the financial support of the COMET K2 - Competence Centres
for Excellent Technologies Programme of the Austrian Federal Ministry for
Transport, Innovation and Technology (BMVIT), the Austrian Federal Ministry of
Science, Research and Economy (BMWFW), the Austrian Research Promotion
Agency (FFG), the Province of Styria and the Styrian Business Promotion Agency
(SFG). We also acknowledge the support of the COST action TU1105.
4D.5 References
[1] Mallikarjuna and T. Kant, A critical review and some results of recently
developed refined theories of of fiber-reinforced laminated composites and
sandwiches. Composite Structures, 23, 293–312, 1993.
[2] H. Denli and J. Sun, Structural-acoustic optimisation of composite sandwich
structures: A review. The Shock and Vibration Digest, 39 (3), 189–200, 2007.
[3] P. Yang, C. Norris and Y. Stavsky, Elastic wave propagation in heterogeneous
plates. International Journal of Solids and Structures, 2, 665–684, 1966.
[4] S. Yu and W. Cleghorn, Free flexural vibration analysis of symmetric honeycomb
panels. Journal of Sound and Vibration, 284, 189–204, 2005.
______________________________________________________________________________________
274
[5] T.-S. Lok and Q.-H. Cheng, Free and forced vibration of simply supported,
orthotropic sandwich panel. Computers and Structures, 79, 301–312, 2001.
[6] D. Mead, A comparison of some equations for the flexural vibration of damped
sandwich beams. Journal of Sound and Vibration, 83 (3), 363–377, 1982.
[7] H. Zhou and G. Li, Free vibration analysis of sandwich plates with laminated
faces using spline finite point method. Computers and Structures, 59 (2), 257–263,
1996.
[8] J. Banerjee, C. Cheung, R. Morishima, M. Perera and J. Njuguna, Free vibration
of a threelayered sandwich beam using the dynamic stiffness method and
experiment. International Journal of Solids and Structures, 44, 7543–7563, 2007.
[9] P. Gaudenzi and R. Carbonaro, An evaluation of higher-order effects on the
eigenfrequencies of composite structures. Composite Structures, 47, 821–825,
1999.
[10] T. Kant and K. Swaminathan, Analytical solutions for free vibration of laminated
composite and sandwich plates based on a higher-order refined theory. Composite
Structures, 53, 73–85, 2001.
[11] D. Backström, Vibration of sandwich beams. Doctoral Thesis, KTH Stockholm,
Sweden, 2006.
[12] E. Nilsson and A. Nilsson, Prediction and measurement of some dynamic
properties of sandwich structures with honeycomb and foam cores. Journal of
Sound and Vibration, 251 (3), 409–430, 2002.
[13] Y. Sadasiva Rao and B. Nakra, Vibrations of unsymmetrical sandwich beams
and plates with viscoelastic cores. Journal of Sound and Vibration, 34 (3), 309–
326, 1974.
[14] M. Maheri, R. Adams and J. Hugon, Vibration damping in sandwich panels.
Journal of Material Science, 43, 6604–6618, 2008.
[15] L. Cremer, M. Heckl, B. Petersson, Structure-borne sound: Structural vibrations
and sound radiation at audio frequencies, 3rd edition. Spiringer, Berlin
Heidelberg, Germany, 2005.
[16] E. Reissner, The effect of transverse shear deformation on the bending of elastic
plates. Journal of Applied Mechanics, 12 (2), A69–A77, 1945.
[17] J. Whitney, Shear correction factors for orthotropic laminates under static load.
Journal of Applied Mechanics, 40, 302–303, 1973.
[18] C. Bert, Simplified analysis of static shear factors for beams of nonhomogeneous
cross section. Journal of Composite Materials, 7, 525–529, 1973.
[19] S. Vlachoutsis, Shear correction factors for plates and shells. International
Journal for Numerical Methods in Engineering, 33, 1537–1552, 1992.
[20] P. Madabhusi-Raman and J. Davalos, Static shear correction factor for laminated
rectangular beams. Composites: Part B, 27B, 285–293, 1996.
______________________________________________________________________________________
275
______________________________________________________________________________________
276
Annex 4E
NVH Simulation of Electric Drivetrains: SRM Case
Fabio Santos, Cassio Faria, Herman Van der Auweraer

Siemens Industry Software, Interleuvenlaan 68, 3001 Leuven, Belgium
Herman.vanderauweraer@siemens.com
Abstract. CAE-based optimization of electric drivetrains is a challenging task due to the

interaction between electromagnetic, structural and acoustic effects. The electric motor noise
generation problem is briefly reviewed and an integrated multi-physics approach is presented
and applied to the case of a switched reluctance based drivetrain system. The baseline
functional model that is used to calculate the structural loads for an FE-based NVH analysis of
the motor is then used in a multiscale thermal model of the electric vehicle into which the
motor is integrated. Two simulation examples illustrate the practical usability of the methods,
one focusing on the NVH aspects and the other on the energy management.
4E.1 Introduction
In an effort to build more fuel-efficient With the electrification of road vehicles,

several new design engineering challenges emerge. New components and subsystems
such as battery and electric drive systems are introduced and performance parameters
such as energy consumption get a new meaning because of the intrinsic range
implications. The specific nature of these components and their vehicle integration
requirements impose to critically investigate the design engineering process itself and
explore new tools and approaches. Below, a CAE process for the electric motor is
discussed.
Currently, the focus for designing electric drives primarily is on energy efficiency.
Carmakers increasingly realize that this must be complemented by including the
acoustic behaviour. Whereas indeed overall noise levels can expected to be lower
than the ones produced by internal combustion engines (ICEs), the sound quality is
often compromised due to the high tonal nature of the generated noise, the effect of
which can be aggravated by coinciding structural resonances. Vibro-acoustic
simulation and analysis early on in the design cycle is a must to prevent these from
happening, and to deliver customers high-quality products.
Specifically for reluctance machines, which offer an interesting drive system
alternative due to their rare-earth free and mechanically simple design, the higher
level of torque ripple significantly increases the acoustic emissions, which further
aggravates the “annoyance” perception by the users. To address this issue and avoid
corrective measures later in the product development process, it is important to
incorporate vibro-acoustic analysis early on in the design process.
______________________________________________________________________________________
277
4E.2 Multiphysics modelling of an EM drivertrain
The multiphysics nature of the NVH problem is obvious. The primary drive principle
relies on harmonic electromagnetic forces which are generated when rotor and stator
are unaligned. These forces have a radial and a tangential component, as illustrated in
Fig. 1-E. For operational purposes, the tangential component is the most interesting
one, as it produces the power.
Fig. 1-E. Electromagnetic forces as generated in a SRM.
The radial component is absorbed in the rotor and the stator, and transformed in an
elastic deformation. These forces may lead to structural vibrations, which finally
produce a pressure field that is radiated from the motor outer surface to the
surroundings or is transmitted through the connections with the supporting structure.
This process is illustrated in Fig. 2-E, showing the interaction between the different
physical domains.
Fig. 2-E. Multiphysics vibro-acoustic model of an electric motor
The modelling and simulation of such a complex process hence requires using a set of
different tools and lots of expertise in the multiple fields to appropriately capture the
phenomena. A very important characteristic that, to some extent, simplifies the
problem at hand is the weak two-way coupling between the domains, i.e. there is a
unidirectional flow of “information” between domains. For example, while the
______________________________________________________________________________________
278
vibration of the structure produces pressure waves in the air surrounding the motor,
these same waves have little or no effect on the vibration of the structure.
Different approaches can be followed to build the multiphysics simulation model.
Through a connection of Electromagnetic and structural Finite Element models, the
transient EM forces can be procesed to structural responses for each relative position
of rotor and stator. This allows to estimate the structural and vibro-acoustic response
of the stator. Important considerations relate to the correct functional connection of
system variables, but also the correct connectivity between the electromagnetic and
the structural meshes. For example, the EM mesh has a higher refinement close to the
air gap because the phenomenon there dominates the system dynamics, meanwhile
the refinement for the structural mesh is concentrated near the smaller section of the
structure. (Fig. 3-E).
Fig. 3-E. Direct comparison between two example meshes used for electromagnetic and
structural simulations
Alternatively, through the use of a 1-D system model, the electromagnetic position
dependent forces can furthermore be processed to time domain torque and torque
ripple profiles and current and efficiency values. The resulting multilevel modeling
approach can then be used to study structural but also thermal/energetic system
responses and even the motor controller design. The model can furthermore be
integrated in a complete vehicle model that includes other components of the
powertrain such as the battery, a combustion engine in case of a HEV, and also other
energy consumers and heat producers, like HVAC, the inverter, and cabin thermal
models. In this way, driving cycles and scenarios can be simulated such that a global
power consumption prediction can be performed, cooling systems can be designed
and typical motor operating conditions can be derived.
The same detailed functional model can also be used to calculate the phase currents
which, using an electromagnetic FEM model, yield the radial stator forces at each
rotor position (and time step). After transformation to the frequency domain and
application to a structural model, surface accelerations and hence acoustic
contributions can be calculated, leading to a noise, vibration and harshness (NVH)
______________________________________________________________________________________
279
model. Consequently, the result of the defined modeling approach is that the design
parameters for the motor, the cooling system and the motor control can be linked to
multiple vehicle performance parameters. Fig.4-E shows the overall scheme of this
modeling approach.
Fig. 4-E. Multiscale SR motor model for full vehicle integration
The discussion will further focus on the latter approach and is illustrated for the
specific case of the design and vehicle integration of a Switched-Reluctance (SR)
electric motor. Such SR-based drive systems gain interest because of their robust and
low-cost designs and the absence of the need of rare earth materials. To develop such
motors, it is a prerequisite to take into account from the start not only component
performance specifications, but also the vehicle integration and vehicle performance
requirements.
4E.3 The Switched Reluctance Motor
The switched reluctance motor (SRM) is a type of synchronous machine, but with
particular features - the stator windings make use of field coils, but no coil or
magnetic material is present on the rotor. The motor works by energizing two
opposing stator poles, generating a magnetic field. This magnetic field forces the
rotation of the closest rotor poles to the position of minimum reluctance, aligning
them to the stator poles. Fig. 5-E shows the energizing and alignment of the stator and
rotor poles. By energizing consecutive stator poles, continuous rotation is generated
______________________________________________________________________________________
280
on the rotor. The SRM functionality and modeling is thoroughly explained in [1]. The
switched reluctance motors are usually referred to by their number of stator and rotor
poles - hence, Fig. 5-E represents an 8/6 SRM (8 stator poles and 6 rotor poles). The
way that each phase is energized, also referred to as turn-on and turn-off angles or
firing angles, is an important part of the motor activation. The proper synchronization
of the phase currents with the rotor position minimizes torque ripple [2] and helps
reduce current peaks and consequently copper losses.
Fig. 5-E. Switched reluctance motor functionality
4E.4 Motor Functional Model
The switched reluctance motor is a non-linear system mainly due to the non-linear
inductance profile, leading to a non-linear torque equation [3]. Analytical solutions
for modeling of the SRM exist [4], but a more detailed and precise model can be
obtained by the use of look-up tables containing the non-linear inductance profile and
other important magnetic characteristics. Look-up tables give the best results overall
and for this reason they were chosen as the method to represent the 1D 12/8 SRM
system. To obtain the magnetic characteristics for this virtual model, an
electromagnetic finite element model can be used [5] such as the freeware Finite
Element Method Magnetics (FEMM) software [6] or the commercial software JMAG.
Such model uses the stator and rotor geometry and material magnetic properties, as
well as the coil characteristics such as the number of wire turns, to obtain the steady
state system response (Fig. 6-E), from which magnetic flux and torque can be
extracted for a given rotor angular position and current.
______________________________________________________________________________________
281
Fig. 6-E. SRM electro-magnetic finite element modeling - (a) stator and rotor geometry; (b)
magnetic flux plot
This procedure is repeated obtaining flux and torque for a discrete range of current
and rotor position values that cover the whole operating range of the SRM [7]. Since
each phase individually has the same behavior for a given current and relative rotor
position and assuming that magnetic flux superposition is true for all cases, the look-
up table data can be extracted only for a single phase and replicated for the other ones.
Once the look-up tables are obtained, the complete functional model can be
implemented, including models for controls, power electronics, loads, activation
angles and losses. The 1-D functional model was developed in the software LMS
Imagine.Lab AMESim [8], [9], a 1-D multi-domain simulation software that allows
for interactions between mechanical and electrical domains and is very suitable for
the simulation of FEV and HEV, integrating different physical domains. Fig.7-E
shows the 12/8 SRM system with 3 half-bridge converters and activation angle
control.
Fig. 7-E. 1-D functional model of a 12/8 SRM with half-bridge converters and controlled firing
angles
______________________________________________________________________________________
282
4E.5 NVH Model
The acoustic contribution of the electric motors on FEV’s and HEV’s must be
predicted as early as possible in the design phase. The dominant cause of noise
generated from the electric motors is considered to be the stator vibrations caused by
radial forces [11][11]. The radial magnetic forces in the air gap between rotor and
stator teeth excite vibrations that are transfered through the structure to the external
surface of the motor and through the air.
Basically, the SRM noise prediction can be divided in 3 consecutive parts: (1)
calculation of stator radial forces; (2) vibration or structure-borne transmission of the
forces; (3) noise generation. By using the detailed 1-D model described above, one
can calculate the phase currents and motor rotor position of a given working or load
condition. This data is used in the 2-D magnetic FE model of the motor to determine
the magnetic forces in the air gap acting on the stator teeth. These forces are surface
density and volume density magnetic forces, but the latter can be neglected [12]. This
procedure is done for only one full rotation, given that the motor is working on steady
state conditions. The resulting forces are then transformed to the frequency domain.
Subsequently, the structural modes of the stator are calculated by means of finite
element method. Fig. 8-E shows the first 10 structural vibration modes of the 12/8
SRM stator. These vibration modes are used together with the magnetic forces
generated previously to calculate a modal-based forced response, which consists of
the surface accelerations of the stator based on the magnetic forces excitation.
Finally, the calculated accelerations are the boundary conditions to be used with the
Boundary Element Method (BEM) direct method, used for exterior problems.
Fig. 8-E. Ten first vibration modes of a 12/8 switched reluctance motor stator
The whole process of the acoustic prediction methodology is summarized in Fig.9-E.

Each step of shows the system inputs that can be modified. In the first step, the most
important inputs are the material and coil magnetic properties, as well as the stator
______________________________________________________________________________________
283
and rotor geometry. For the 1-D simulation, the most important parameters are the
different control strategies that can be implemented, together with different load
conditions and/or operation cycles. The third step of the process is where the force
calculations take place, and on the last part of the process, the most important inputs
are the structural properties of the motor stator, which can affect directly the sound
pressure levels around the motor.
Fig. 9-E. Acoustic prediction diagram
4E.6 Energy Model
The investigation and simulation of the global energy consumption of the FEV’s and
HEV’s is an important stage in the development of these vehicles [13]. By estimating
the different system losses it is possible to dimension adequately the vehicle’s
systems, such as motor and cabin cooling or battery power consumption, based on the
operating conditions. Similarly, the vehicle powertrain characteristics can be
evaluated before prototyping, based on suitable driving cycle scenarios. To obtain a
computationally efficient simulation system for the global energy flow, a simpler
motor model has to be used, providing the simulation with just the necessary data
needed for the losses computations. This behavioural model contains data that can be
extracted from a more detailed motor model, as discussed above, or from real
measurements.
Three look-up tables are used in this simpler model: (1) Maximum motor torque in
function of input voltage, rotary velocity and temperature, (2) Maximum generator
torque in function of input voltage, rotary velocity and temperature, (3) Motor losses
in function of torque, rotary velocity and temperature. The tables containing the
maximum motor and generator torque can easily be obtained using the detailed
functional model by running it under different conditions of rotary velocity, input
voltage and temperature. The average torque is extracted from the system and the
______________________________________________________________________________________
284
tables can be created. To obtain the losses table, a submodel was created, dividing the
losses in 3 different types - copper losses, core losses and mechanical losses [14].
Copper losses are generated when the current goes through the coil, iron losses are
related to the magnetic flux that passes through iron (stator and rotor) and mechanical
losses are generated from friction. Thereafter the 3 tables can be utilized in the SRM
behavioural model that is included in a global electric vehicle system specially
designed to verify the system’s energy flow. The complete energy flow model of this
electric vehicle is seen in Fig. 10-EErreur ! Source du renvoi introuvable..
Fig. 10-E. AMESim model of an electric vehicle with energy flow management
The Driver, ECU and Vehicle subsystems are used to simulate braking and
acceleration commands, as well as simple translational dynamics for the whole
vehicle, given a constant load. The Vehicle subsystem also includes an air-
conditioning model that is able to regulate the cabin temperature. Benchmark driving
cycles consisting of input velocities are used to simulate different operating
conditions.
The switched reluctance motor and its thermal subsystems - inverter, battery and the
SRM itself - are used to compute the temperature and energy losses in the different
component parts. To model the temperature distribution in the SRM, a lumped
thermal model of the SRM was used based on a thermal equivalent circuit for the
temperature distribution in the motor. The motor is divided into several thermal parts:
frame, stator yoke, stator teeth, rotor core, rotor teeth, coil windings and lateral
______________________________________________________________________________________
285
windings, axial shaft, bearings, internal air and air gap [15]. The whole thermal
network is shown in Fig. 11-E.
Fig. 11-E. SRM thermal network

The thermal component of the SRM also allows to connect the cooling system. Two
inputs that are connected directly to the frame and stator yoke allow to create a
thermal network representing the cooling system and the effect on the motor
temperature can be visualized on the motor itself. The inverter losses are modeled as a
half-bridge with two types of components - diodes and transistors - and the losses are
also divided into two types - conduction and switching losses [16]. By using the
motor rotational velocity as input, the switching losses are calculated, and the
conduction losses are calculated based on peak and average current values.
Losses on the battery are estimated with a model that takes into account the internal
resistance of the battery, as well as the open circuit voltage.
4E.7 Simulation Examples
Two simulations were carried out to verify both the NVH model of the SRM and the
vehicle thermal model of the electric vehicle.
Acoustic Analysis
Following the procedure described above, an acoustic prediction of the SRM was
carried out for two working conditions. The firing angles which are very important in
the SRM with respect to maximum and minimum current and torque production were
chosen as parameters to be investigated in view of their effect on the sound pressure
levels produced by the motor. The two cases chosen were: firing angles optimized to
maximize torque and firing angles optimized to maximize torque and minimize mean
current levels. The objective was to evaluate if, by minimizing the coil current, a
significant improvement on the acoustic quality of the SRM could be realixed, due to
the reduction of the magnetic field amplitude and hence the forces exciting the stator
teeth. The parameters used for the 1-D and 3-D simulations are shown in Table 1-E.
To control the torque generated by the motor, a simple PI controller was used, and the
______________________________________________________________________________________
286
chosen rotary velocity was guaranteed by attaching an auxiliary motor to the SRM
shaft.
Table 1-E. Acoustic Simulation Specifications
Parameter Value
Rotational speed 6000 RPM
Torque 30 Nm
Minimum frequency 200 Hz
Maximum frequency 3500 Hz
A Boundary Element sound prediction method was used, allowing the use of field
point meshes, to visualize the system’s spatial response. Fig. 12-E shows the sound
pressure levels of the two simulation cases for a planar field point mesh of 1.2m by
1.2m at 1604 Hz.
Fig. 12-E. Comparison of the sound pressure levels of the two simulation cases at 1604 Hz - (a)
SRM with firing angle optimization with regards to torque (b) SRM with firing angle
optimization with regards to torque and current
To assess the acoustic response over the whole frequency range, a spherical field
point mesh was used to obtain the average value of the acoustic power over all points
around the stator. Fig. 13-E shows the acoustic power responses of both simulations.
______________________________________________________________________________________
287
Fig. 13-E. Comparison of the acoustic power response of a 12/8 SRM at 6000 RPM and 30 Nm
torque - two firing angle strategies: (–) System response with maximized torque in respect to
the firing angles ( - - ) System response with maximized torque and minimized current with
respect to the firing angles
The simulation results show a clear difference between the acoustic properties in the
two cases. The minimization of the current has an important effect on the acoustic
power response of the SRM. However, on some frequency ranges the system with
current minimization has higher acoustic power levels. This happens because the
current levels at that particular frequency band are higher, even though the mean
current levels are lower for that case.
Electric Vehicle Energy Management Simulation

To evaluate and test the global vehicle model, two simple cooling scenarios were
! !
used, one with cooling fluid temperature at 35 C and one at 45 C. Both scenarios
consider the use of a hypothetical cooling jacket to cool the SRM frame. The
objective is to observe the motor temperature with respect to the cooling fluid
temperatures. This sort of procedure can be very helpful on determining the optimal
specifications of the cooling circuit of the motor, so that on one hand the motor
temperature is kept within operational limits, and on the other hand no over-
dimensioning of the cooling circuit is done. The simulations use the New European
Driving Cycle (NEDC) as the operational conditions. This driving cycle contains both
an urban driving cycle and an extra-urban driving cycle, relating to city and highway
driving modes, respectively, with total duration of 1180 seconds. The vehicle velocity
(Km/h) in function of the total driving cycle time for the NEDC is shown in Fig. 14-
EErreur ! Source du renvoi introuvable..
______________________________________________________________________________________
288
Fig. 14-E. New European Driving Cycle used for the global electric vehicle simulation, with
urban and highway cycles
The main motor and system losses (total SRM losses, copper, iron and mechanical
losses, battery and inverter losses) for both simulation scenarios are shown in Fig. 15-
E.
Fig. 15-E. Integrated electric vehicle system losses simulation results. Left: total SRM losses,
iron losses, mechanical losses. Right: Battery and inverter losses
The temperature rise in the SRM is mainly caused by the heating of the copper
windings. Consequently, the windings and stator teeth are the most critical part to be
monitored in the SRM. To verify the effectiveness of the cooling circuit, the
temperature rise was observed on the two critical parts plus on the stator yoke, which
is in direct contact with the frame and the cooling circuit. Erreur ! Source du renvoi
introuvable. 16-E shows the temperature rise for both simulation scenarios - cooling
! !
fluid and jacket at 45 C and at 35 C
______________________________________________________________________________________
289
Fig. 16-E. Temperature rise on a NEDC cycle, analyzed at critical points of the Switched
Reluctance Motor - stator yoke, stator tooth and coil windings. Left: results for cooling fluid of
45°, right cooling fluid of 35°
Analyzing both simulation scenarios, the difference of the temperature values in the
SRM parts can be clearly seen. The temperature on the coil windings is well
influenced by the cooling circuit.
4E.8 Conclusion
A multiphysical system model was developed for studying the acoustic and thermal
behaviour of an electromotor. A detailed functional model was created by means of
an integrated approach that uses magnetic finite element modeling to generate look-up
tables that contain the non-linear inductance profile and magnetic torque
characteristics. These tables are the basis for a 1-D simulation model that yields the
torque, phase currents and rotor position. Basedon this model, two applications were
realized: an NVH prediction approach to compute sound pressure levels around the
motor and a global electric vehicle system model to compute losses and thermal
management.
The modeling approach was applied to the case of an SRM motor. It was shown how
the control setting of the motor can affect the radiated sound. The structural
parameters of the stator are also relevant, and can be modified to verify the benefits in
the system. Furthermore, trade-offs can be nvestigated so that the optimal parameters
can be chosen based on system specifications. In a second application, the
behavioural motor model was used in an energy management system calculation.
Using the detailed functional model, peak values that could otherwise go unchecked
are taken into account, leading to more realistic results. The thermal properties of the
motor can be evaluated, such as the main source of heat and/or losses, and how the
material properties affect the distribution of the energy. Two examples show how the
motor cooling system specifications can be defined based on driving cycle
simulations.
______________________________________________________________________________________
290
Acknowledgement
VLAIO, the Flemish Agency for Innovation, is acknowledged for their financial
support in the SRMOTIF project and Flanders’ Drive for their financial support with
the Electric Powertrain project.
4E.9 References
[1] R. Krishnan. Switched Reluctance Motor Drives: Modeling, Simulation. CRC

Press, 2001.
[2] C. Mademlis and I. Kioskeridis. Gain-scheduling regulator for high-performance
position control of switched reluctance motor drives. Industrial Electronics, IEEE
Transactions on, 57(9):2922–2931, 2010.
[3] S.M. Lukic and A. Emadi. State-switching control technique for switched
reluctance motor drives: Theory and implementation. Industrial Electronics, IEEE
Transactions on, 57(9):2932–2938, 2010.
[4] W. Ding and D. Liang. A fast analytical model for an integrated switched
reluctance starter/generator. Energy Conversion, IEEE Transactions on,
25(4):948–956, 2010.
[5] K. Mhatli and B. Ben Salah. Improved modeling of switched reluctance motor
including mutual and saturation effects. In MELECON- 15th IEEE Mediterranean
Electrotechnical Conference, pages 1470–1475. IEEE, 2010.
[6] H. Kuss, T. Wichert, and B. Szvmanski. Design of a high speed switched
reluctance motor for spindle drive. In Compatibility in Power Electronics, 2007.
CPE’07, pages 1–5. IEEE, 2007.
[7] N.H. Fuengwarodsakul, R.W. De Doncker, and RB Inderka. Simulation model of
a switched reluctance drive in 42 v application. In Industrial Electronics Society,
2003. IECON’03. The 29th Annual Conference of the IEEE, volume 3, pages
2871–2876. IEEE, 2003.
[8] R. Meillier, T. Rozier, A. Brizard, and D. Almer. Optimization of vehicle energy
management through multi-domain system simulation. In Transport Research
Arena Europe, 2010.
[9] L. Broglia, M. Ponchant, and R. Meillier. Balancing vehicle energy performance
with drivability and thermal comfort: benefits of a multi-domain system
simulation approach in the case of an electric-hybrid vehicle. In 25th World
Electric Vehicle Symposium and Exposition, 2009.
[10] J. Florentin, F. Durieux, Y. Kuriyama, T. Yamamoto, M. DeBruin, S. Jordan, J.
Han, G. Barber, Q. Zou, X. Sun, et al. Electric motor noise in a lightweight steel
vehicle. Analysis, 1:1062, 2011.
[11] J.O. Fiedler, K.A. Kasper, and R.W. De Doncker. Calculation of the acoustic
noise spectrum of srm using modal superposition. Industrial Electronics, IEEE
Transactions on, 57(9):2939–2945, 2010.
[12] M. van der Giet, E. Lange, DAP Correa, IE Chabu, SI Nabeta, and K. Hameyer.
Acoustic simulation of a special switched reluctance drive by means of field–
circuit coupling and multiphysics simulation. Industrial Electronics, IEEE
Transactions on, 57(9):2946–2953, 2010.
______________________________________________________________________________________
291
[13] V.P. Atluri, K. Koprubasi, and N.D. Brinkman. Analytical evaluation of
propulsion system architectures for future urban vehicles. In Advanced Hybrid
vehicle Powertrains, 2011.
[14] M. Woon, X. Lin, A. Ivanco, A. Moskalik, C. Gray, and Z.S. Filipi. Energy
management options for an electric vehicle with hydraulic regeneration system. In
Advanced Hybrid vehicle Powertrains, 2011.
[15] H. Rouhani, J. Faiz, and C. Lucas. Lumped thermal model for switched
reluctance motor applied to mechanical design optimization. Mathematical and
computer modelling, 45(5-6):625–638, 2007.
[16] MH Bierhoff and FW Fuchs. Semiconductor losses in voltage source and current
source igbt converters based on analytical derivation. In Power Electronics
Specialists Conference, 2004. PESC 04. 2004 IEEE 35th Annual, volume 4, pages
2836–2842. IEEE, 2004.
______________________________________________________________________________________
292
Annex 4F
CAE Approach to Electric Vehicle Warning Sound System Design
Bert Van Genechten, Koen Vansant, Herman Van der Auweraer

Siemens Industry Software, Interleuvenlaan 68, 3001 Leuven, Belgium
Herman.vanderauweraer@siemens.com
Abstract. Electric vehicles need to be equipped with a sound generation system to warn
vulnerable road users of their presence, in particular at low speeds. Next to generating the right
sound, also the actual sound system needs to be designed and configured in order to have a
maximal effect. A numerical simulation approach was developed and applied to the case of
bumper integration of a steerable sound source.
4F.1 Introduction
A topic that is increasingly considered as a critical noise performance for

electric/hybrid vehicles is the relation between the exterior sound of quiet road
vehicles and the safety of Vulnerable Road Users (VRU). In particular at low speeds,
before the tire noise becomes observable, the absence of any perceived engine noise
and hence the absence of any vehicle proximity warning, may cause danger to other
road users.
Perceptual considerations for the sound design [1], [2], [3] lead to considerations
regarding harmonic and narrow/broadband spectral content, speed dependency and
modulations. To be relevant at potential VRU listener positions, the propagation of
the sound from source to receiver has to be considered, taking in to account the
influences of the surrounding structure and space (car body, road, road obstacles, road
layout…) and taking into account the masking effect from the traffic noise (Fig 1-F).
A proper system configuration needs to identify the number of sources, the source
characteristics and the optimal source locations.
A numerical simulation approach was hetero investigated and applied in the design of
a transducer array to be installed on the bumper of the eVADER vehicle [1], [2].
eVADER was an EU funded research project in the FP7 framework in which the
problem of warning sound generation to protect vulnerable road users was
investigated, leading to the installation of a prototype system on a demonstrator
vehicle. Such simulation models allow assessing more realistically the impact of the
system integration on the performance of the warning system [4]. Using the acoustic
responses obtained with these models, the strategy selected ifor the source design is
verified. Next, a sensitivity study is presented to assess the impact of the most critical
changes in the vehicle’s acoustic environment: temperature changes, changes in the
road surface properties and the influence of nearby scattering objects such as e.g.
parked vehicles.
______________________________________________________________________________________
293
Fig. 1-F. Vehicle Warning Sound design challenges and synthesis model
4F.2 CAE Tools for exterior warning sound design
Simulation Approach
Whenever virtual prototypes are used for design support or performance evaluation,
the selection of a suitable discretization approach and analysis type is key. When
considering acoustic radiation problems, which involve solving an uncoupled exterior
acoustic problem with imposed normal velocity boundary conditions in the frequency
domain two families of numerical models are frequently used in the current
engineering practice. The Boundary Element Method (BEM) is based on a boundary
discretization and is hence very suited to model acoustic radiation.
Moreover, significant advances such as the Fast Multipole BEM methods [5], [6], [7]
and the Hierarchical Matrix (or H-Matrix) solvers [8] have been made in the BEM’s
computational efficiency. The Finite Element Method (FEM) has become the standard
solution approach for solving interior acoustic and vibro-acoustic problems. Different
strategies for coping with this limitation have been devised such as infinite elements
and perfectly matched layer (PML) approaches [9] applied on fictitious outer FEM
boundaries. Based on the strengths and weaknesses of the different modelling
approaches a combination of different strategies is proposed for this design study. For
scenarios where only the bumper itself needs to be considered, the FEM-AML
technique which is based on the PML approach is selected.
Integration Study
A detailed geometrical model of the front bumper of the eVADER demonstrator
vehicle is incorporated in the numerical models in this design study. The front of the
car is approximated by extruding the outer edge of the surface model in the length
direction (Fig. 2-F). Due to their small size, the six individual loudspeakers are
______________________________________________________________________________________
294
incorporated as the six red circular vibrating pistons shown in the left. To finalize the
creation of the FEM-AML model, the pre-processing capabilities of the LMS
Virtual.Lab Acoustics [10] suite of numerical tools are used to automatically generate
a suitable convex wrapper mesh around the extended bumper geometry. The right side
of Fig. 2-F shows the final baseline FEM-AML model with an average mesh size such
that it is at least valid up to 1.3kHz according to commonly used rules of thumb. The
total computational cost for solving this baseline FEM model of 120.000 degrees of
freedom is 9 seconds per frequency. In comparison, solving an equivalent H-matrix
BEM model takes 210 seconds while providing similar acoustic predictions.
Fig. 2-F. (left) Computational bumper geometry, (right) FEM-AML model of the warning
signal generator
Fig. 3-F. Acoustic pressure responses obtained with the baseline FEM-AML model: (left)
individual acoustic transfer functions for the six speakers, (right) forward-facing acoustic beam
at 900Hz
The FEM-AML model of the warning signal generator installed on the front bumper
is used in two steps:
• Firstly to obtain the acoustic transfer functions between the surface velocity of
each individual loudspeaker and the acoustic pressure at a number of semi-
circular microphone arrays. These transfer functions are used as an input for the
beam forming strategies. An example of such transfer functions is shown in the
left of Fig 3-F.
• Secondly, by introducing the optimal beam forming control parameters in the
numerical model, the acoustic pressure generated by the warning generator can
be reconstructed. This information is useful to assess the impact the warning
generator has on the environment noise landscape.
______________________________________________________________________________________
295
Design Sensitivity Study
Given the high level of exposure of the exterior warning system to a range of
environmental conditions it is important that the dominant parameters are taken into
account during the design of the system to ensure robustness. Based on a literature
review the following parameters have been found to influence the propagation of
acoustic waves in typical traffic environments: (i) Variations in the acoustic properties
of the air that may be caused by the presence of temperature gradients or variations in
relative air humidity, (ii) Aerodynamic disturbances of the propagating medium
caused by wind gradients and air turbulence, (iii) Impact of second order reflections
that may interfere with the direct field generated by the array such as the presence of
reflecting objects and ground reflections and (iv) Interference with other noise sources
that can be other sound sources in the environment or incoherent background noise. In
the current study, the first and third class of influencing factors were considered. The
fourth impact type is implicitly integrated into the total warning system that
automatically selects a suitable absolute sound level to be generated.
Fig. 4-F. Impact of temperature and humidity on speed of sound and density (upper) and on
beam forming (lower)
Both the ambient temperature and the air’s relative humidity change the acoustic
wave propagation properties and hence may significantly impact the warning
system’s performance. In order to identify the most influential parameter, Fig. 4-F
shows the evolution of both density and sound speed for representative temperature (-
10oC to +30oC, blue curves) and humidity (0 to 100%, red curves) ranges. The curves
clearly show that the temperature has the most impact. Since temperature changes
______________________________________________________________________________________
296
only impact the acoustic material properties of the ambient fluid, the selected FEM-
AML technique remains valid. By combining these updated models with the control
coefficients computed for the reference conditions, the robustness with respect to
ambient fluid property changes is assessed. The right side of Fig. 5-F shows a
representative example result of this study for a warning signal targeted at a 0o
orientation at 20oC and at 0oC. A very good agreement between the obtained
directivities can be observed.
A second parameter with a major impact on the warning system’s performance is the
acoustic absorption of the road surface. Apart from attenuating acoustic waves, the
acoustic impedance of the road surface may also introduce asymmetry or even fully
scatter the warning signal. In order to incorporate these effects a description of the
frequency-dependent acoustic normal impedance is incorporated in a discretization of
the road surface. In [11] a theoretical model was developed in which the acoustic
surface impedance depends on the ambient fluid properties and also on the air flow
resistivity and porosity of the asphalt. Typical values for these quantities were taken
from literature. Since the exterior warning system is designed to warn at-risk VRUs
up to large distances in front of the vehicle, the road surface should be appropriately
modelled. Hence, the area where the road surface is located is modelled explicitly for
a scenario where the vehicle is driving along a 20m long straight road which is two
lanes wide, as is shown in the left of Fig. 5-F. Additionally the impact that scattering
objects in the vicinity of the warning signal generator is considered. To this end, the
BEM model is extended by adding a model of two vehicles parked on the side of the
road 5m in front of the approaching demonstrator vehicle. The vehicles themselves
are considered to be perfectly reflecting and their additional impact on the spatial
distribution and directivity of the acoustic warning signals will be assessed. The
increased extent and complexity of the problem setting significantly impacts the
choice of the modelling approach. Based on the model size (around 190.000 degrees
of freedom for the road surface and 250.000 when including the parked cars) the H-
matrix BEM was selected. Due to the increased computational cost (around 13-25 min
per frequency of interest), the acoustic responses are only computed every 25Hz.
The right side of Fig. 5-F shows the spatial distribution of the acoustic pressure
amplitude generated by the warning system at 600Hz for a steering angle of 0o. This
case is considered as the worst possible mode of operation since the central acoustic
pressure amplitude lobe of the warning signal impinges directly on both stationary
vehicles. As can be seen in the contour plots, the presence of the two vehicles shields
the areas behind them as expected. The sound waves that are scattered by the vehicles
do interfere with the central lobe of the warning signal but do not alter its shape nor
do they generate areas of very low SPL.
______________________________________________________________________________________
297
Fig. 5-F. Impact of road surface impedance and nearby scattering objects on beam forming
performance: (left) H-matrix BEM model discretisation, (right) spatial distribution of the
acoustic pressure field at 600Hz
The conclusions above for the spatial pressure distributions at discrete frequencies are
confirmed by the directivity analysis presented in Fig. 6-F. On the left side the
directivity obtained for the baseline problem setting is given. This is compared to the
directivity for a 0o steering where only the road surface impedance (middle) and
where both the road surface and the two parked vehicles (right) are included in the
analysis. In both cases, the directivity pattern becomes asymmetric due to the
asymmetric positioning of the vehicle with respect to the road (it occupies the right
lane). Additionally the interference between the direct radiated acoustic field and the
acoustic waves scattered by the parked vehicles is clearly visible in the image on the
right but as was observed in Fig. 5-F these interferences do not distort the overall
directionality and hence they do not impact the system’s intended operation.
Fig. 6-F. Impact of road surface impedance and nearby scattering objects on beam forming
performance: (left) reference configuration, (middle) impact of 20m long asphalt road, (right)
impact of asphalt road and parked cars
4F.3 CAE Tools for interior warning sound design
Additional to warning the vulnerable road users, also the driver needs to be alerted of
the presence of such road users. Detection systems such as pedestrian detection and
obstacle warning systems become available and their input can be used to generate an
appropriate warning signal to the driver. Such systems should provide clear
information about the diagnosed risk, allowing the driver to take appropriate action.
______________________________________________________________________________________
298
In this design, both acoustic and visual warnings are merged to achieve optimal driver
alertness and to support his or her decision making. For the acoustic diver warning,
the use of the existing vehicle audio system to provide intuitively correct sensational
information on the detected threat situation was explored through a simulation-based
acoustic configuration analysis approach towards internal warning systems, more
specifically, to investigate the generation of direction feeling on the interior warning
sound so as to optimally provide information on the detected direction of the threat
situation (generation of a so-called “phantom source” correlated to the VRU). These
simulations which are based on Ray Tracing and binaural auralization have provided
inputs regarding optimal speaker configuration and control (including optimal use of
already available audio systems through pre-filtered acoustic warning signals). The
digital filter design process is shown schematically in Fig. 7-F.
Fig. 7-F. Simulation-based interior warning system design.
4F.4 Conclusion
A CAE-based analysis and evaluation is proposed to support the design of the an

acoustic exterior warning system. In this process, advanced acoustic beam forming
algorithms are linked to high-fidelity numerical models for the assessment of the
acoustic radiation of the different transducers as installed on the vehicle. In this way,
a thorough assessment of different configurations can be executed in a flexible and
efficient manner. The choice of modelling technique is based on the problem setting
and the expected computation times for different CAE models. Based on the control
coefficients for a localized warning signal targeted at different angles with respect to
the forward direction of travel, a verification study of the spatial distribution of the
acoustic pressure was performed. Finally it was demonstrated that also the design of
an internal warning system for the driver can be designed using CAE tools, in
particular Ray Tracing methods.
______________________________________________________________________________________
299
4F.5 References
[6] D. Quinn, J. Garcia, J. Mitchell, P. Clark, "eVADER : the development and

evaluation results of a next-generation pedestrian alert solution for quiet electric
vehicles", Fisita World Automotive Congress 2014 (Maastricht, the Netherlands,
June 2-6 ,2014
[7] J. J. García , J. M. Barrios , D. Quinn , G. Baudet ,T. Schmit , J. Chamard, “The
expected impact of eVADER technology on the safety of vulnerable road users”,
Proc. ISMA 2014, Noise and Vibration Conference (Leuven, Belgium, 15-17
September 2014
[8] H. Van der Auweraer, D. Sabbatini, E. Sana, K. Janssens, “Analysis of Electric
Vehicle Noise in View of VRU Safety”, in Proceedings ICSV18, Brazil, 10-14
July 2011
[9] Van Genechten B., Vansant K. & Berkhoff A.P., Simulation-based design of a
steerable acoustic warning device to increase (H)EV detectability while reducing
urban noise pollution In Proceedings of the TRA 2014, Paris, France, 14-17 April
2014
[10] R. Hallez, K. De Langhe, “Solving large industrial acoustic models with the fast
multipole method”, Proc. ICSV 16, Krakow (Poland), 2009
[11] K. Vansant, R. Hallez, K. De Langhe, H. Bériot, G. Massa, S. Donders, H. Van
der Auweraer, “Simulating acoustic engine performance over a broad frequency
range”, SAE Paper 2011-26-0019, Proc. SIAT 2011, Pune, India, Jan. 19-21,
2011
[12] Gumerov N.A. and Duraiswami R., Fast Multipole Methods for the Helmholtz
equation in three dimensions, Elsevier Science, 2005
[13] Hackbusch, W., “A sparse matrix arithmetic based on H-matrices. Part I:
introduction to H-matrices”, Computing, 62(2):89–108, 1999.
[14] Bérenger, J., “A perfectly matched layer for the absorption of electromagnetic
waves”, Journal of Computational Physics, 114:157–171, 1994
[10] LMS Virtual.Lab Rev 12
[11] J.F. Hamet and M. Bérengier, “Acoustical characteristics of porous pavements: a
new phenomenological model,” Proc. Internoise 93 (Leuven, Belgium) 641-646
(1993)
______________________________________________________________________________________
300
Annex 4G
Damage detection in the inertial dampers of the electrical
vehicle using experimental modal identification
Calin Itu1, Paul Nicolae Borza1, Maria Luminita Scutaru1, Consuelo Gragera Pena2,
Sofia Sanchez Mateo3, Sorin Vlase1
1
TRANSILVANIA University of Brașov, Romania
2
Extremadura University, Badahoz, Spain
3
Mechanical Engineering and Energy Department, Miguel Hernandez University of
Elche, Av de la Universidad, S/N - 03202, Elche, Spain,
4G.1 Introduction
Inertial dampers ensure, through the structural design, mechanical energy dissipation
due to vibration or noise generating source. Also these dampers allow, as result of
vibration energy dissipation in volume of plastics on the one hand and as result of
dissipative higher power efficiency and stability on the other hand (due to the high
energy absorption distributed system) the use of corresponding nonlinear zone
flexibility based plastics. Moreover, the spatial distribution of inertial masses ensure a
more uniform thermal loading plastic base and can be achieved without breaking
conditions thereby avoiding pregnancy and the development of concentrators zones in
basic plastic. The same spatial distribution in conjunction with inertial masses form,
their surface roughness and especially the flexibility of plastics allow obtaining
anisotropic features extremely useful in terms of optimal adaptation in relation to the
type of shock absorber damping designed for this task. In [11] such a solution is
presented that allows vibration damping electronic control circuit due to the
commutation of a stepper motor power. If evoked system consists of silicone plastic
housed in a hermetic housing, and moving inertial masses is limited mechanically. If
plastic studied system consists of a silicone housed in a hermetic housing and the
motion of the inertial masses is limited mechanically.
In [1] are shown the theoretical equations of motion corresponding inertial mass
immersed in basic plastic. In the paper [2] such dampers are included in the class of
'meta-materials', mechanical hybrid structures respectively (composite) that allow
"predicting the mechanical properties of the design phase" and damping properties in
relation to specific needs in system which the damper is included. Such systems are
found in the living world that can fulfil both dissipation function but also
amplification and therefore accelerations of vibration transmitted through the damper.
These issues are extremely important in case of electric vehicles, as a result their
shareholders through either the power of synchronous motors or asynchronous motors
with the control based on the electric commutation circuit occurring two categories of
sources of vibration or noise frequency bands with two distinct areas. First, lower,
______________________________________________________________________________________
301
corresponding to the fundamental frequency applied to the electric motor that is
dependent on variable angular speed.
The second frequency range of vibration is due to electrical switchgear and circuit
characteristics and correspond to higher frequencies such as those produced by
electrical switching circuit. This second field will vary very little in relation to the
angular speed of the engine or the velocity of the vehicle. This second area is intended
to be covered by the damping characteristics (band pass filter) that can perform
inertial damper.
Alternative solutions, involving the development of an annulator of oscillations in
electric commutation type Wiener filter [8], which both in terms of reliability, energy
consumption, especially high computing power necessary to implement satisfy only
partially the needs of applications involving a rotating as uniform for electric vehicle
drive system. The Wiener filter operation is based on fundamental control signal
adding an estimated signal so that the resulting signal at the output compensate the
vibrations due to the commutation electric power applied to the electric motor. The
use of a computer system in real time requires high computing power or specialized
circuits but also requires a certain energy and so strictly administered in case of an
electric vehicle. Furthermore, if for some reason practically disappears the power
control system of the torsional vibration, disappears too the possibility to decrease the
vibrations.
4G.2 Mechanical model and analysis scenarios
In Fig 1-G a standard model of an inertial damper is presented. The damper is made
by rubber reinforced with balls. In the paper two cases are considered: the balls are
made by steel and, alternately, by aluminium (Al). The results are similar, which is
why we only present the case where steel balls are. The properties for rubber and steel
are those currently used in engineering.
Fig. 1-G. Physical model of the inertial damper
A finite element model for the damper is presented in Fig.2-G. The main goal of the
paper is to determine the eigenvalues for a functional structure and for the damaged
______________________________________________________________________________________
302
structure. Making an analysis in the Eigen frequencies changes is possible to identify
the moment when the damage is produced. The method was presented in previous
paper as [3],[4],[7]. Analytical methods in the field of the eigenvalues can be found in
many paper (see [5],[6]).
Fig. 2-G.. Finite element model of the damper
Fig. 3-G. Frequency response for a rubber model
______________________________________________________________________________________
303
Fig. 4-G. Frequency response for an inertial damper with steel balls
Fig. 5-G. Damage scenarios for the damper
______________________________________________________________________________________
304
Fig. 6-G. Damage scenarios for the damper
Fig. 7-G. Scenarios of analysis
4G.2 Eigenvalues for the damaged damper
The eigenvalues and eigen modes were computed using a finite element soft.
The result for the eigenvalues computed for the scenarios presented in Fig. 7-G are
presented in Table 1-G.
Table 1.G. Eigenvalues for the first 10 scenarios
______________________________________________________________________________________
305
Mode no. Frequency [Hz]
Good (1) D1 (2) D2 (3) D3 (4) D4 (5) D5 (6) D6 (7) D7 (8) D8 (9) D9 (10)
1 320 320 320 305 318 302 212 312 280 170
2 320 321 321 311 320 310 243 318 305 223
3 459 460 458 437 456 438 383 451 427 349
4 837 835 814 554 824 711 461 822 730 500
5 879 874 859 727 860 852 711 863 835 671
6 879 876 864 843 880 858 844 872 854 803
7 1379 1356 1245 934 1366 1270 1009 1381 1383 1375
8 1559 1541 1460 1140 1547 1487 1232 1559 1555 1495
9 1560 1554 1519 1289 1562 1515 1369 1561 1560 1505
10 1600 1600 1588 1382 1602 1590 1548 1600 1592 1570
A representation of the first ten eigenvalues for the scenarios considered is made in
the Fig. 8-G to Fig. 12-G.
Fig. 8-G. The first two eigenvalues for the ten scenarios
Fig. 9-G. The 3rd and 4th eigenvalues for the ten scenarios
______________________________________________________________________________________
306
Fig. 10-G. The 5th and 6th eigenvalues for the ten scenarios
Fig. 13-G. The first 10 eigenvalues for the scenario nr. 10 (initial and damaged model)
______________________________________________________________________________________
307
Symmetric modes gives equal eigenvalues [9], [10] but a crack in the material makes
one of the eigen modes to keep his value while the other will change. That is why it is
necessary to consider various eigen modes each time to make comparison. Some
eigenvalues, despite the emergence of crack and damage of the damper, do not change
(it can compare, for example, line 1 and 2 of Table 1-G).
If we analyse one of the cases, for example, the last case (version 10 – Fig. 13-G) you
can see that the eigen mode 7, for example, is not changed although the case study
shows a consistent fissure. In contrast to most other eigenvalues are severely affected.
It will result so that some eigenvalues are not altered by a change in material
continuity while others give some consistent differences. It follows therefore that for
detecting a beginning of damage is necessary to consider several vibration modes to
be comparable to the initial situation.
Table 2-G. Eigenvalues for the next 10 scenarios (balls made of aluminium)
Mode no. Frequency [Hz]
Good D1 D2 D3 D4 D5 D6 D7 D8 D9
(11) (12) (13) (14) (15) (16) (17) (18) (19) (20)
1 327 327 326 312 325 309 217 318 286 174
2 327 328 328 317 326 316 248 325 311 228
3 467 467 465 445 464 446 390 459 434 356
4 855 853 831 565 841 725 471 839 745 511
5 895 891 876 741 877 870 725 879 851 684
6 895 893 881 861 897 875 861 888 871 819
7 1401 1379 1265 951 1389 1291 1029 1404 1406 1398
8 1592 1574 1489 1164 1579 1516 1258 1592 1588 1526
9 1592 1587 1550 1315 1594 1546 1394 1594 1593 1536
10 1634 1634 1621 1408 1637 1624 1580 1634 1626 1605
When we use balls made by aluminium the results are presented in Table 2-G. The
conclusions analysing the values presented in Table 2-G offer similar conclusions as
for the values presented in Table 2-G, from a qualitative point of view. The density of
aluminium being smaller as the density for steel, the quantitative results differ.
______________________________________________________________________________________
308
4G.3. Conclusions
In the operation of a vibration damper, insert mode of the balls inertial in the mass
damper rubber makes it vulnerable to damage. It is, of course, be very useful to detect
early failures of this damper before damage. It can thus avoid significant material
expenses. A periodic visual inspection requires expenses and waste of time. For this
reason it can be very useful to find a method of early detection of cracks in the rubber
body. In this paper we propose the use of their eigen frequencies damper
measurement and comparison to baseline. If there are significant differences the
damper may be replaced. As seen in the presented figures, some eigen frequencies can
differ more or less, depending on the place and dimension of the fissure. For this
reason, multiple frequencies should be analysed and compared simultaneously. We
must not confine ourselves to the study of a single frequency of failure because there
may be cases where the eigen frequency corresponding to a particular mode of
vibration, not be influenced by the mechanism of destruction, but other frequencies
are affected.
The analysis was carried out for a single shock absorber, attached at one end. In
practice it may be more useful to consider the damper mounted in the mechanical
system. This arrangement makes its eigen frequencies of the damper to change. They
can be determined from a more complex calculus but the findings of this study do not
change. From the results presented it can be concluded that a change of eigen
frequency more than 5% can pull the alarm on the possibility of cracks in the structure
while a difference of more than 10% should lead to a revision of the whole assembly
to determine the cause of this change.
4G.4 References
[1].Carotti,A., Turci,E., A tuning criterion for the inertial tuned damper. Design using
phasors in the Argand–Gauss plane. Applied Mathematical Modelling, 23(3),
199–217, 1999.
[2].Claeys,C. C., Pluymers,B., Sas,P., Desmet,W., Design of a resonant metamaterial
based acoustic enclosure, Proceedings of ISMA2014, 3351–3358, 2014.
[3].Gillich,G.-R., Praisach,Z.-I., Iavornic,C.M., Reliable method to detect and assess
damages in beams based on frequency changes. ASME 2012 International Design
Engineering Technical Conferences and Computers and Information in
Engineering Conference. pp. 129-137, 2012.
[4].Gillich,G.-R., Praisach,Z.-I., Damage-patterns-based method to locate
discontinuities in beams. SPIE Smart Structures and Materials - Nondestructive
Evaluation and Health Monitoring. 869532-8, 2013.
[5].Neubert, V.H., Design of Damping and Control Matrices by Modification of
Eigenvalues and Eigenvectors, Shock and Vibration, Volume 1, Issue 4, Pages
317-329, 1994.
______________________________________________________________________________________
309
[6].Nobari, A.S., Ewins, D.J., On the Effectiveness of Using Only Eigenvalues in
Structural Model Updating Problems, Shock and Vibration, Volume 1, Issue 4,
Pages 339-348, 1994.
[7].Praisach,Z.-I., Gillich,G.-R., Birdeanu,D.E., Considerations on natural frequency
changes in damaged cantilever beams using FEM. International Conference on
Engineering Mechanics, Structures, Engineering Geology, International
Conference on Geography and Geology-Proceedings. pp. 214-219, 2010.
[8].Shimkin,N., Estimation and Identification in Dynamical Systems. Israel Institute
of Technology, Department of Electrical Engineering, Lecture Notes, 2009,
Chapter 3, Fall 2009.
[9].Vlase,S., Danasel,C., Scutaru,M.L., Mihalcica,M., Finite Element Analysis of a
Two- Dimensional Linear Elastic Systems with a Plane "Rigid Motion",
Romanian Journal of Physics; 59(5-6), 2014.
[10]. Vlase,S., Teodorescu,P.P., Itu,C., Scutaru,M.L., Elasto-Dynamics of a
Solid with a General "Rigid" Motion Using FEM Model. Part II. Analysis of a
Double Cardan Joint. Romanian Journal of Physics, 58 (7-8), 2013.
[11]. ***, "Inertial Damper for Stepper Motor" Focus, I. (2012). Dmp 20 / 29 / 37,
42. Phytro Corp. see on www.phytron.com/DMP April 2015.
______________________________________________________________________________________
310

Ch4 Numerical Approaches For The NVH Study of Electric and Hybrid Electric Vehicles

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ch4 Numerical Approaches For The NVH Study of Electric and Hybrid Electric Vehicles

Uploaded by

Copyright:

Available Formats

NVH Analysis Techniques for Design and

Optimization of Hybrid and Electric Vehicles

Paul Borza1 and Sofia Sanchez Mateo 2

2 Methodologies and numerical contributions to the study of

4A. Vibro-acoustic metamaterial developments combining low-mass

FEM modelling for

BEM approach to the

Some important achievements have been presented in this chapter, as well as

In this chapter an innovative metamaterial concept to improve acoustic behaviour is

4A.2 Metamaterial creation by inclusion of resonant structures

4A.3 Stop band prediction

Name Symbol Value

To prove the potential of the introduced metamaterials concept to reduce acoustic

Fig. 9-A: Pictures of the demonstrator

4A.5 Measurement results

Further investigation of the produced demonstrator explains the shift in frequency;

4B.6.1 Model of the test set-up

The wooden plate is modelled as a baffle: an infinite plane which is acoustically

To compute the acoustic radiated power through the enclosure a BE modal-based

A numerical simulation of the demonstrator with this geometry of the resonator

4B.1 Motion equations of an one-dimensional finite element

4B.1.1 General considerations

4B.1.2 Lagrangian of the Mechanical System

where the vector of nodal displacements δe is:

If v and w are the displacements of a bar point in the directions Oy and Oz

or, with respect to the global reference system:

The kinetic energy expression for the whole bar is:

where, for the sake of simplicity I xx = I x , I yy = I y , I zz = I z . The angular velocity

w = N (w )δe ; v = N (v )δe Eq. 16-B

The deformation energy due to torsion is:

The inertia force is given by:

Fig. 2-B. Determination of the axial inertia force

⎧µ x ⎫ ⎧ X!!o ⎫ ⎧1⎫ ⎧1⎫

We obtain the internal energy due to inertia in the form of:

The external work of distributed loads is:

and that of concentrated loads q e TL in the nodes is:

The Lagrangian for the element is[42][43][44][45][46] and [47]:

4B.2 Two-dimensional finite element

4B.2.1 Motion’s equations for a two-dimensional finite element

Fig. 3-B. A two-dimensional finite element with plan motion

If the M point is subjected to a displacement f L , transforming into the M' point, we

global reference system. The continuous displacement field f (x,y) L is

f L = N (x,y) δ e (t) L Eq. 33-B

σ = D·ε Eq. 37-B

where, for a two-dimensional finite element (plane state of deformation) we have:

ε=a· f Eq. 39-B

where a is a differentiation operator:

Considering all this, for the deformation energy we have:

with k e being the stiffness matrix:

The nodal forces q T

The Lagrangian for the considered element is:

We can write the equations of motion in concentrated form:

m e δ!! e, L + 2 c e δ! e, L + [ k e + k e (ε) + k e (ω2)] δ e, L =

c e (ω)= ∫ N T RT R! N ρdV ; k e (ε) + k e (ω2) = ∫N

represents the angular velocity operator corresponding to the angular velocity ω. We

The equations of motion will be, in this case:

4B.2.2 Modal analysis of a rotating disk

Table 1-B. Eigenvalues of a rotating disk

Fig. 5-B. Eigenmodes 7, 8 and 9

Fig. 6-B. Eigenmodes 10,11 and 12

Fig. 7-B. Eigenmodes 13,14 and 15

Fig. 9-B. Eigenmodes 7, 8 and 9

Fig. 10-B. Eigenmodes 10, 11 and 12